Index

A

ABC AbstractPolicyWorker (class in dockyard_rl.models.policy.workers.base_policy_worker) accessibility_tree (dockyard_rl.experience.cua.backends.interface.Observation attribute) accumulate_grads() (in module dockyard_rl.models.jax.train_step) accuracy (dockyard_rl.algorithms.dpo.DPOValMetrics attribute) (dockyard_rl.algorithms.rm.RMValMetrics attribute) accuracy() (in module dockyard_rl.algorithms.dpo_evals) ACK (dockyard_rl.models.policy.utils.IPCProtocol attribute) ack_timeout_ms (dockyard_rl.data_plane.interfaces.DataPlaneConfig attribute) acquire() (dockyard_rl.experience.cua.provisioner.DockerPoolProvisioner method) (dockyard_rl.experience.cua.provisioner.Provisioner method) (dockyard_rl.experience.cua.provisioner.StaticPoolProvisioner method) act() (dockyard_rl.experience.cua.backends.control_center.ControlCenterBackend method) (dockyard_rl.experience.cua.backends.interface.OSWorldBackend method) (dockyard_rl.experience.cua.backends.macos.MacBackend method) (dockyard_rl.experience.cua.backends.official.OfficialDesktopEnvBackend method) (dockyard_rl.experience.cua.backends.windows.WindowsBackend method) action_history (dockyard_rl.experience.cua.vendor.desktop_env_eval.grading_env.GradingEnv attribute) actor_class_fqn (dockyard_rl.environments.utils.EnvRegistryEntry attribute) add() (dockyard_rl.algorithms.async_utils.interfaces.ReplayBufferProtocol method) (dockyard_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) add_bos (dockyard_rl.data.DataConfig attribute) add_eos (dockyard_rl.data.DataConfig attribute) add_generation_prompt (dockyard_rl.data.DataConfig attribute) add_grpo_token_loss_masks_and_generation_logprobs() (in module dockyard_rl.algorithms.grpo) add_loss_mask_to_message_log() (in module dockyard_rl.data.llm_message_utils) add_ref_logprobs_to_data() (in module dockyard_rl.algorithms.dpo) add_system_prompt (dockyard_rl.data.DataConfig attribute) adv_estimator (dockyard_rl.algorithms.grpo.GRPOConfig attribute) advantage_clip_high (dockyard_rl.algorithms.grpo.GRPOConfig attribute) advantage_clip_low (dockyard_rl.algorithms.grpo.GRPOConfig attribute) advantages (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossDataDict attribute) AdvEstimatorConfig (class in dockyard_rl.algorithms.grpo) aggregate_rollout_metrics() (in module dockyard_rl.algorithms.grpo) aggregate_spec_decode_counters() (in module dockyard_rl.models.generation.vllm.utils) AIME2024Dataset (class in dockyard_rl.data.datasets.response_datasets.aime24) AIMEDataset (class in dockyard_rl.data.datasets.eval_datasets.aime) AIMEEvalDataConfig (class in dockyard_rl.data) algorithm (dockyard_rl.distributed.batched_data_dict.SequencePackingArgs attribute) alias (dockyard_rl.models.policy.teacher_scorer.TeacherScorer attribute) (dockyard_rl.models.policy.teacher_worker_group.TeacherConfig attribute) align() (dockyard_rl.algorithms.x_token.token_aligner.TokenAligner method) align_routed_expert_indices() (in module dockyard_rl.models.generation.vllm.router_capture) alignment_from_flat_batch() (in module dockyard_rl.algorithms.x_token.loss_utils) AlignmentBatch (class in dockyard_rl.algorithms.x_token.token_aligner) AlignmentPair (class in dockyard_rl.algorithms.x_token.token_aligner) all_to_all_sq2vp() (in module dockyard_rl.distributed.model_utils) all_to_all_vp2sq() (in module dockyard_rl.distributed.model_utils) allgather_cp_sharded_tensor() (in module dockyard_rl.distributed.model_utils) AllGatherCPTensor (class in dockyard_rl.distributed.model_utils) allocate_rollout_budget() (in module dockyard_rl.algorithms.rollout_allocation) allocation_exponent (dockyard_rl.algorithms.routing_diversity.RoutingDiversityConfig attribute) allow_auto_truncate (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) allow_multimodal_inputs (dockyard_rl.models.generation.interfaces.GenerationConfig attribute) AllTaskProcessedDataset (class in dockyard_rl.data.datasets.processed_dataset) AllToAllDispatchMetadata (class in dockyard_rl.models.dtensor.moe.dispatch) (class in dockyard_rl.models.jax.moe.alltoall) AllToAllTokenDispatcher (class in dockyard_rl.models.dtensor.moe.dispatch) (class in dockyard_rl.models.jax.moe.alltoall) alpha (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) answers (dockyard_rl.environments.interfaces.EnvironmentReturn attribute) api_token (dockyard_rl.environments.code_environment.CodeEnvConfig attribute) (dockyard_rl.environments.gdpval_agentic_environment.GDPvalAgenticEnvConfig attribute) (dockyard_rl.environments.program_bench_environment.ProgramBenchEnvConfig attribute) (dockyard_rl.environments.terminal_bench_environment.TerminalBenchEnvConfig attribute) apply_effort_reward() (in module dockyard_rl.algorithms.effort_reward)

apply_fp8_patches() (in module dockyard_rl.models.generation.vllm.quantization.fp8) apply_invalid_action_penalty() (in module dockyard_rl.algorithms.reward_functions) apply_message_span_advantage_penalties() (in module dockyard_rl.algorithms.reward_functions) apply_moe_surgery() (in module dockyard_rl.models.dtensor.moe.surgery) apply_reward_shaping() (in module dockyard_rl.algorithms.reward_functions) apply_rules_to_timeFormat() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.misc) apply_top_k_top_p() (in module dockyard_rl.algorithms.logits_sampling_utils) are_lists_equal() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.utils) arguments (dockyard_rl.tool_protocol.hermes.ParsedToolCall attribute) arguments_json (dockyard_rl.tool_protocol.hermes.ParsedToolCall attribute) Array (in module dockyard_rl.models.jax.layers) (in module dockyard_rl.models.jax.linear_attn.delta_rule) (in module dockyard_rl.models.jax.linear_attn.gated_deltanet) (in module dockyard_rl.models.jax.loss) (in module dockyard_rl.models.jax.models.gemma4) (in module dockyard_rl.models.jax.models.qwen3) (in module dockyard_rl.models.jax.models.qwen3_next) (in module dockyard_rl.models.jax.moe.alltoall) (in module dockyard_rl.models.jax.moe.block) (in module dockyard_rl.models.jax.moe.dispatch) (in module dockyard_rl.models.jax.moe.experts) (in module dockyard_rl.models.jax.moe.load_balance) (in module dockyard_rl.models.jax.moe.qwen3_moe) (in module dockyard_rl.models.jax.moe.router) (in module dockyard_rl.models.jax.train_step) ArrayLike (in module dockyard_rl.models.generation.vllm.router_capture) artifact_location (dockyard_rl.utils.logger.MLflowConfig attribute) as_dict() (dockyard_rl.models.policy.interfaces.Timer method) as_tensor() (dockyard_rl.data.multimodal_utils.PackedTensor method) assemble_teacher_logits_from_shards() (in module dockyard_rl.algorithms.x_token.loss_utils) assert_finite_sampling_params() (in module dockyard_rl.models.generation.interfaces) assert_no_double_bos() (in module dockyard_rl.data.datasets.utils) assert_prev_logprobs_available() (in module dockyard_rl.algorithms.opd) assert_teacher_student_batch_grid() (in module dockyard_rl.algorithms.x_token.utils) assert_valid() (in module dockyard_rl.experience.cua.tasks.validate) assert_xtoken_ipc_node_local() (in module dockyard_rl.algorithms.x_token.utils) assess_assistant_turn() (in module dockyard_rl.rewards.invalid_action) assess_tool_calls() (in module dockyard_rl.tool_protocol.registry) assign_group_ids() (in module dockyard_rl.algorithms.rollout_allocation) async_engine (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) async_generate_response_for_sample_turn() (in module dockyard_rl.experience.rollouts) async_grpo (dockyard_rl.algorithms.grpo.GRPOConfig attribute) async_grpo_train() (in module dockyard_rl.algorithms.grpo) AsyncGRPOConfig (class in dockyard_rl.algorithms.grpo) AsyncLoopThread (class in dockyard_rl.models.generation.sglang.utils) AsyncTrajectoryCollector (class in dockyard_rl.algorithms.async_utils.trajectory_collector) AsyncTrajectoryCollectorImpl (class in dockyard_rl.algorithms.async_utils.trajectory_collector) attach_message_log_view() (in module dockyard_rl.data.llm_message_utils) attempted (dockyard_rl.tool_protocol.hermes.HermesExtraction property) (dockyard_rl.tool_protocol.registry.ToolUseAssessment attribute) attention_backend (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) attention_bias (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) (dockyard_rl.models.jax.models.qwen3.Qwen3Config attribute) attention_k_eq_v (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) AUTOMODEL_FACTORY (in module dockyard_rl.models.policy.utils) available_algorithms() (in module dockyard_rl.data.packing.algorithms) avg_sequences_per_pack (dockyard_rl.data.packing.metrics.PackingMetrics property) (dockyard_rl.data.packing.packing.PackingMetrics property) avg_tokens_per_sequence (dockyard_rl.data.packing.metrics.PackingMetrics property) (dockyard_rl.data.packing.packing.PackingMetrics property) AVQADataset (class in dockyard_rl.data.datasets.response_datasets.avqa) AXIS_CP (in module dockyard_rl.models.jax.sharding) AXIS_DP (in module dockyard_rl.models.jax.sharding) AXIS_EP (in module dockyard_rl.models.jax.sharding) AXIS_TP (in module dockyard_rl.models.jax.sharding)

B

backend (dockyard_rl.data_plane.interfaces.DataPlaneConfig attribute) (dockyard_rl.models.generation.interfaces.GenerationConfig attribute) (dockyard_rl.tool_protocol.protocol.StructuredToolUseConfig attribute) backfill_routed_experts() (in module dockyard_rl.data.llm_message_utils) backward() (dockyard_rl.algorithms.x_token.loss_utils.Fp32SparseMM static method) (dockyard_rl.distributed.model_utils.AllGatherCPTensor static method) (dockyard_rl.distributed.model_utils.ChunkedDistributedEntropy static method) (dockyard_rl.distributed.model_utils.ChunkedDistributedGatherLogprob static method) (dockyard_rl.distributed.model_utils.ChunkedDistributedHiddenStatesToLogprobs static method) (dockyard_rl.distributed.model_utils.ChunkedDistributedLogprob static method) (dockyard_rl.distributed.model_utils.ChunkedDistributedLogprobWithSampling static method) (dockyard_rl.distributed.model_utils.DistributedCrossEntropy static method) (dockyard_rl.distributed.model_utils.DistributedLogprob static method) (dockyard_rl.distributed.model_utils.DistributedLogprobWithSampling static method) base_commit (dockyard_rl.sandbox.task_executor_client.TaskSpec attribute) base_penalty_for_type() (in module dockyard_rl.rewards.invalid_action) BaseMathEnvironment (class in dockyard_rl.environments.math_environment) BaseVllmGenerationWorker (class in dockyard_rl.models.generation.vllm.vllm_worker) batch_multiplier (dockyard_rl.algorithms.grpo.GRPOConfig attribute) batch_size (dockyard_rl.distributed.batched_data_dict.BatchedDataDict property) (dockyard_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) batched_message_log_to_flat_message() (in module dockyard_rl.data.llm_message_utils) BatchedDataDict (class in dockyard_rl.distributed.batched_data_dict) bbox_giou_reward() (in module dockyard_rl.environments.rewards) begin_episode() (dockyard_rl.experience.cua.environment.OSWorldEnvironment method) bf16 BinaryPreferenceDataset (class in dockyard_rl.data.datasets.preference_datasets.binary_preference_dataset) bind_router_replay() (in module dockyard_rl.models.dtensor.moe.router_replay) block_network (dockyard_rl.environments.gdpval_agentic_environment.GDPvalAgenticEnvConfig attribute) (dockyard_rl.sandbox.task_executor_client.SessionStartSpec attribute) (dockyard_rl.sandbox.task_executor_client.TaskSpec attribute) BM25Retriever (class in dockyard_rl.environments.retriever) BOOK (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.table) both_resident_bytes (dockyard_rl.algorithms.distillation_memory.MemoryReport property) boxed (in module dockyard_rl.environments.rewards) brier_score() (in module dockyard_rl.algorithms.dpo_evals)

broadcast() (dockyard_rl.distributed.stateless_process_group.StatelessProcessGroup method) broadcast_weights_for_collective() (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.interfaces.ColocatablePolicyInterface method) (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) Buffer (class in dockyard_rl.models.jax.moe) build_abstention_pair() (in module dockyard_rl.algorithms.rlaif) build_args (dockyard_rl.sandbox.task_executor_client.SessionStartSpec attribute) build_cluster() (in module dockyard_rl.cluster.fleet) build_constitutional_pair() (in module dockyard_rl.algorithms.rlaif) build_context (dockyard_rl.sandbox.task_executor_client.SessionStartSpec attribute) build_data_plane_client() (in module dockyard_rl.data_plane.factory) build_exact_token_map() (in module dockyard_rl.algorithms.x_token.loss_utils) build_extraction_command() (in module dockyard_rl.environments.gdpval_agentic_environment) build_grader() (in module dockyard_rl.experience.cua.grading.factory) build_memory_report() (in module dockyard_rl.algorithms.distillation_memory) build_mesh() (in module dockyard_rl.models.jax.sharding) build_moe_block() (in module dockyard_rl.models.jax.moe.qwen3_moe) build_moe_mesh() (in module dockyard_rl.models.jax.sharding) build_moe_meshes() (in module dockyard_rl.models.dtensor.moe.mesh) build_optimizer() (in module dockyard_rl.models.jax.optimizer) build_preference_loss() (in module dockyard_rl.algorithms.loss.loss_functions) build_preference_pairs() (in module dockyard_rl.algorithms.online_dpo) build_provisioner() (in module dockyard_rl.experience.cua.provisioner) build_qwen3next_moe_block() (in module dockyard_rl.models.jax.models.qwen3_next) build_read_file_command() (in module dockyard_rl.environments.multi_turn_session_env) build_rlaif_preference_data() (in module dockyard_rl.algorithms.rlaif) build_schedule() (in module dockyard_rl.models.jax.optimizer) build_session_write_command() (in module dockyard_rl.environments.multi_turn_session_env) build_structural_tag() (in module dockyard_rl.models.generation.vllm.utils) build_timeout (dockyard_rl.sandbox.task_executor_client.SessionStartSpec attribute) build_token_dispatcher() (in module dockyard_rl.models.dtensor.moe.config) build_vllm_modelopt_nvfp4_config() (in module dockyard_rl.modelopt.utils) build_write_file_command() (in module dockyard_rl.environments.gdpval_agentic_environment) BuildBatchFn (in module dockyard_rl.algorithms.online_dpo) bytes_outstanding (dockyard_rl.data_plane.observability.DataPlaneStats attribute) bytes_outstanding_by_partition() (dockyard_rl.data_plane.observability.MetricsDataPlaneClient method)

C

cache_dir (dockyard_rl.experience.cua.vendor.desktop_env_eval.grading_env.GradingEnv attribute) calculate_advantages_on_gpu (dockyard_rl.algorithms.grpo.GRPOConfig attribute) calculate_aligned_size() (in module dockyard_rl.models.policy.utils) calculate_baseline_and_std_per_prompt() (in module dockyard_rl.algorithms.utils) calculate_kl() (in module dockyard_rl.algorithms.loss.utils) (in module dockyard_rl.models.jax.loss) calculate_pass_rate_per_prompt() (in module dockyard_rl.environments.metrics) calculate_rewards() (in module dockyard_rl.experience.rollouts) calibrate_qkv_fp8_scales() (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.interfaces.PolicyInterface method) (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) callback (dockyard_rl.data_plane.interfaces.ObservabilityConfig attribute) called_workers (dockyard_rl.distributed.worker_groups.MultiWorkerFuture attribute) Candidate (class in dockyard_rl.algorithms.online_dpo) CandidateGroups (in module dockyard_rl.algorithms.online_dpo) candidates_per_prompt (dockyard_rl.algorithms.online_dpo.OnlineDPOConfig attribute) canonical_token() (in module dockyard_rl.algorithms.x_token.token_aligner) cap_drop (dockyard_rl.sandbox.task_executor_client.SessionStartSpec attribute) capture_diff (dockyard_rl.sandbox.task_executor_client.TaskSpec attribute) cast_tensor_to_fp8_blockwise() (in module dockyard_rl.models.generation.vllm.quantization.fp8) causal_conv1d() (in module dockyard_rl.models.jax.linear_attn.delta_rule) cDPO CDPOLossFn (class in dockyard_rl.algorithms.loss.loss_functions) ce_label_mask() (in module dockyard_rl.algorithms.x_token.loss_utils) ce_loss_scale (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) chat() (dockyard_rl.algorithms.rlaif.ChatLLM method) (dockyard_rl.rewards.hle_grader.HLEJudgeClient method) chat_template_tools() (dockyard_rl.tool_protocol.registry.ToolRegistry method) ChatLLM (class in dockyard_rl.algorithms.rlaif) check_accessibility_tree() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) check_consumption_status() (dockyard_rl.data_plane.adapters.noop.NoOpDataPlaneClient method) (dockyard_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) (dockyard_rl.data_plane.interfaces.DataPlaneClient method) (dockyard_rl.data_plane.observability.MetricsDataPlaneClient method) check_csv() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) check_direct_json_object() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) check_gnome_favorite_apps() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.basic_os) check_health() (dockyard_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollectorImpl method) check_include_exclude() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) check_json() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) check_line_number() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) check_list() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) check_moved_jpgs() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.basic_os) check_pdf_pages() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.pdf) check_refit_names() (in module dockyard_rl.models.dtensor.moe.refit) check_save() (dockyard_rl.utils.timer.TimeoutChecker method) check_text_enlarged() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.basic_os) checkpoint_dir (dockyard_rl.utils.checkpoint.CheckpointingConfig attribute) checkpoint_must_save_by (dockyard_rl.utils.checkpoint.CheckpointingConfig attribute) checkpoint_path (dockyard_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) checkpointing (dockyard_rl.algorithms.distillation.MasterConfig attribute) (dockyard_rl.algorithms.dpo.MasterConfig attribute) (dockyard_rl.algorithms.grpo.MasterConfig attribute) (dockyard_rl.algorithms.rm.MasterConfig attribute) (dockyard_rl.algorithms.sft.MasterConfig attribute) CheckpointingConfig (class in dockyard_rl.utils.checkpoint) CheckpointManager (class in dockyard_rl.utils.checkpoint) chosen_key (dockyard_rl.data.PreferenceDatasetConfig attribute) chunk_average_finalize() (in module dockyard_rl.algorithms.x_token.loss_utils) chunk_average_log_probs() (in module dockyard_rl.algorithms.x_token.loss_utils) chunk_list_to_workers() (in module dockyard_rl.environments.utils) chunk_log_prob_sums() (in module dockyard_rl.algorithms.x_token.loss_utils) chunk_teacher_logits_for_cross_cluster() (in module dockyard_rl.algorithms.x_token.loss_utils) chunked_prefill_size (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) ChunkedDistributedEntropy (class in dockyard_rl.distributed.model_utils) ChunkedDistributedGatherLogprob (class in dockyard_rl.distributed.model_utils) ChunkedDistributedHiddenStatesToLogprobs (class in dockyard_rl.distributed.model_utils) ChunkedDistributedLogprob (class in dockyard_rl.distributed.model_utils) ChunkedDistributedLogprobWithSampling (class in dockyard_rl.distributed.model_utils) CISPO claim_meta() (dockyard_rl.data_plane.adapters.noop.NoOpDataPlaneClient method) (dockyard_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) (dockyard_rl.data_plane.interfaces.DataPlaneClient method) (dockyard_rl.data_plane.observability.MetricsDataPlaneClient method) claim_meta_poll_interval_s (dockyard_rl.data_plane.interfaces.DataPlaneConfig attribute) cleanup() (dockyard_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) clear() (dockyard_rl.algorithms.async_utils.interfaces.ReplayBufferProtocol method) (dockyard_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) clear_logger_metrics() (dockyard_rl.models.generation.interfaces.GenerationInterface method) (dockyard_rl.models.generation.vllm.vllm_generation.VllmGeneration method) clear_samples() (dockyard_rl.data_plane.adapters.noop.NoOpDataPlaneClient method) (dockyard_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) (dockyard_rl.data_plane.interfaces.DataPlaneClient method) (dockyard_rl.data_plane.observability.MetricsDataPlaneClient method) clear_visible_devices() (in module dockyard_rl.utils.device_query) CLEVRCoGenTDataset (class in dockyard_rl.data.datasets.response_datasets.clevr) clip_grad_by_total_norm_() (in module dockyard_rl.models.dtensor.parallelize) clipped_pg_loss() (in module dockyard_rl.models.jax.loss) ClippedPGLossConfig (class in dockyard_rl.algorithms.loss.loss_functions) ClippedPGLossDataDict (class in dockyard_rl.algorithms.loss.loss_functions) ClippedPGLossFn (class in dockyard_rl.algorithms.loss.loss_functions) close() (dockyard_rl.data_plane.adapters.noop.NoOpDataPlaneClient method) (dockyard_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) (dockyard_rl.data_plane.interfaces.DataPlaneClient method) (dockyard_rl.data_plane.observability.MetricsDataPlaneClient method) (dockyard_rl.experience.cua.clients.control_center.ControlCenterClient method) cluster (dockyard_rl.algorithms.distillation.MasterConfig attribute) (dockyard_rl.algorithms.dpo.MasterConfig attribute) (dockyard_rl.algorithms.grpo.MasterConfig attribute) (dockyard_rl.algorithms.rm.MasterConfig attribute) (dockyard_rl.algorithms.sft.MasterConfig attribute) ClusterConfig (class in dockyard_rl.distributed.virtual_cluster) CNI code (dockyard_rl.experience.cua.actions.pyautogui_parse.ParsedAction attribute) CODE_TOOLS (in module dockyard_rl.tool_protocol.registry) CodeEnvConfig (class in dockyard_rl.environments.code_environment) CodeEnvironment (class in dockyard_rl.environments.code_environment) CodeJaccardEnvConfig (class in dockyard_rl.environments.code_jaccard_environment) CodeJaccardEnvironment (class in dockyard_rl.environments.code_jaccard_environment) CodeJaccardEnvironmentMetadata (class in dockyard_rl.environments.code_jaccard_environment) CodeJaccardVerifyWorker (class in dockyard_rl.environments.code_jaccard_environment) coefficient (dockyard_rl.algorithms.effort_reward.EffortRewardConfig attribute)

collect_overlapping_teacher_shards() (in module dockyard_rl.algorithms.x_token.loss_utils) collection_interval (dockyard_rl.utils.logger.GPUMonitoringConfig attribute) CollectiveWeightSynchronizer (class in dockyard_rl.weight_sync.collective_weight_synchronizer) ColocatablePolicyInterface (class in dockyard_rl.models.policy.interfaces) colocated (dockyard_rl.algorithms.distillation.DistillationConfig attribute) colocation_activation_reserve_gb (dockyard_rl.algorithms.distillation.DistillationConfig attribute) colocation_overhead_gb (dockyard_rl.algorithms.distillation.DistillationConfig attribute) colocation_reserved_num_gpus() (in module dockyard_rl.distributed.worker_groups) colocation_safety_margin (dockyard_rl.algorithms.distillation.DistillationConfig attribute) colocation_schedule (dockyard_rl.algorithms.distillation.DistillationConfig attribute) ColocationSchedule (class in dockyard_rl.algorithms.distillation_memory) ColwiseParallelWithGather (class in dockyard_rl.models.dtensor.parallelize) combine (dockyard_rl.algorithms.opd.TeacherPanelConfig attribute) combine() (dockyard_rl.models.dtensor.moe.dispatch.AllToAllTokenDispatcher method) (dockyard_rl.models.dtensor.moe.dispatch.LocalTokenDispatcher method) (dockyard_rl.models.jax.moe.alltoall.AllToAllTokenDispatcher method) (dockyard_rl.models.jax.moe.dispatch.LocalTokenDispatcher method) COMBINE_MODES (in module dockyard_rl.algorithms.teacher_panel) combine_panel_logprobs() (in module dockyard_rl.algorithms.teacher_panel) combine_reward_functions() (in module dockyard_rl.environments.rewards) COMMON_CHAT_TEMPLATES (class in dockyard_rl.data.chat_templates) compare_conference_city_in_order() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.table) compare_csv() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.table) compare_python_pure_text() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) compare_table() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.table) compare_terminal_and_txt() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) compare_time_in_speedtest_results() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) compare_urls() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.utils) CompilerReward (class in dockyard_rl.rewards.compiler) COMPLETE (dockyard_rl.models.policy.utils.IPCProtocol attribute) compose_deliverable() (in module dockyard_rl.environments.gdpval_agentic_environment) compute_advantage() (dockyard_rl.algorithms.advantage_estimator.GDPOAdvantageEstimator method) (dockyard_rl.algorithms.advantage_estimator.GRPOAdvantageEstimator method) (dockyard_rl.algorithms.advantage_estimator.OPDAdvantageEstimator method) (dockyard_rl.algorithms.advantage_estimator.ReinforcePlusPlusAdvantageEstimator method) compute_and_apply_seq_logprob_error_masking() (in module dockyard_rl.algorithms.grpo) compute_calibration_metrics() (in module dockyard_rl.algorithms.dpo_evals) compute_expert_bias_delta() (in module dockyard_rl.models.dtensor.moe.load_balance) (in module dockyard_rl.models.jax.moe.load_balance) compute_packing_metrics() (in module dockyard_rl.data.packing.metrics) (in module dockyard_rl.data.packing.packing) compute_score() (in module dockyard_rl.environments.dapo_math_verifier) compute_spec_decode_metrics() (in module dockyard_rl.models.generation.vllm.utils) concat() (dockyard_rl.data.multimodal_utils.PackedTensor class method) (dockyard_rl.data_plane.interfaces.KVBatchMeta method) (dockyard_rl.distributed.batched_data_dict.BatchedDataDict static method) configure_dynamo_cache() (in module dockyard_rl.models.policy.utils) configure_generation_config() (in module dockyard_rl.models.generation) configure_rich_logging() (in module dockyard_rl.utils.logger) configure_worker() (dockyard_rl.models.generation.sglang.sglang_worker.SGLangGenerationWorker static method) (dockyard_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker static method) connect() (dockyard_rl.experience.cua.clients.control_center.ControlCenterClient method) Constitution (class in dockyard_rl.algorithms.rlaif) ConstitutionPrinciple (class in dockyard_rl.algorithms.rlaif) constrained_tools (dockyard_rl.models.generation.interfaces.StructuredSpec attribute) (dockyard_rl.tool_protocol.protocol.StructuredToolUseConfig attribute) consumed_samples (dockyard_rl.algorithms.distillation.DistillationSaveState attribute) (dockyard_rl.algorithms.dpo.DPOSaveState attribute) (dockyard_rl.algorithms.grpo.GRPOSaveState attribute) (dockyard_rl.algorithms.rm.RMSaveState attribute) (dockyard_rl.algorithms.sft.SFTSaveState attribute) container_id (dockyard_rl.experience.cua.provisioner.HostLease attribute) content (dockyard_rl.tool_protocol.hermes.HermesExtraction attribute) context_length (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) context_parallel_size (dockyard_rl.algorithms.opd.TeacherResourceConfig attribute) (dockyard_rl.models.jax.config.JaxPolicyConfig attribute) (dockyard_rl.models.policy.teacher_worker_group.TeacherConfig attribute) control (dockyard_rl.experience.cua.actions.pyautogui_parse.ParsedAction attribute) Control (in module dockyard_rl.experience.cua.actions.pyautogui_parse) ControlCenterBackend (class in dockyard_rl.experience.cua.backends.control_center) ControlCenterClient (class in dockyard_rl.experience.cua.clients.control_center) controller (dockyard_rl.experience.cua.vendor.desktop_env_eval.grading_env.GradingEnv attribute) controller_address (dockyard_rl.data_plane.interfaces.DataPlaneConfig attribute) conversation_process_message() (in module dockyard_rl.data.datasets.response_datasets.general_conversation_dataset) conversation_sender_mapping_sample_to_allowed (in module dockyard_rl.data.datasets.response_datasets.general_conversation_dataset) convert_calibration_to_vllm_format() (in module dockyard_rl.models.generation.vllm.quantization.fp8_train_utils) convert_dcp_to_hf() (in module dockyard_rl.utils.native_checkpoint) convert_hf_to_gguf() (in module dockyard_rl.utils.gguf_export) convert_metadata() (in module dockyard_rl.data.datasets.response_datasets.general_conversation_dataset) convert_to_seconds() (in module dockyard_rl.utils.timer) copy() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) copy_defaults() (dockyard_rl.data.interfaces.TaskDataSpec method) CORRECTNESS_KEY (in module dockyard_rl.algorithms.effort_reward) count_moe_blocks() (in module dockyard_rl.models.dtensor.moe.router_replay) CP cp (dockyard_rl.models.dtensor.moe.mesh.MoEParallelDims attribute) (dockyard_rl.models.jax.sharding.MoEMeshDims attribute) cp_load_balanced_to_contiguous() (in module dockyard_rl.distributed.model_utils) cp_shift_next() (in module dockyard_rl.distributed.model_utils) cpu_offload_gb (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) create_context_parallel_ctx() (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl static method) create_env() (in module dockyard_rl.environments.utils) create_teacher_configs_from_opd_config() (in module dockyard_rl.models.policy.teacher_worker_group) create_teacher_worker_groups() (in module dockyard_rl.algorithms.opd) create_weight_synchronizer() (in module dockyard_rl.weight_sync.factory) create_worker() (dockyard_rl.distributed.worker_groups.RayWorkerBuilder.IsolatedWorkerInitializer method) create_worker_async() (dockyard_rl.distributed.worker_groups.RayWorkerBuilder method) criterion (dockyard_rl.rewards.gdpval_rubric.RubricItem attribute) critique() (in module dockyard_rl.algorithms.rlaif) critique_request (dockyard_rl.algorithms.rlaif.ConstitutionPrinciple attribute) CrossTokenizerCollator (class in dockyard_rl.data.cross_tokenizer_collate) CrossTokenizerDistillationLossConfig (class in dockyard_rl.algorithms.loss.loss_functions) CrossTokenizerDistillationLossDataDict (class in dockyard_rl.algorithms.loss.loss_functions) CrossTokenizerDistillationLossFn (class in dockyard_rl.algorithms.loss.loss_functions) cu_seqlens_k (dockyard_rl.models.huggingface.common.FlashAttentionKwargs attribute) cu_seqlens_q (dockyard_rl.models.huggingface.common.FlashAttentionKwargs attribute) CUA cuda_graph_bs (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) cuda_graph_max_bs (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) CUDA_VISIBLE_DEVICES (in module dockyard_rl.utils.device_query) current_epoch (dockyard_rl.algorithms.grpo.GRPOSaveState attribute) current_step (dockyard_rl.algorithms.grpo.GRPOSaveState attribute) custom_dataloader (dockyard_rl.data.DataConfig attribute)

D

DailyOmniDataset (class in dockyard_rl.data.datasets.response_datasets.daily_omni) DAPO DAPOMath17KDataset (class in dockyard_rl.data.datasets.response_datasets.dapo_math) DAPOMathAIME2024Dataset (class in dockyard_rl.data.datasets.response_datasets.dapo_math) data (dockyard_rl.algorithms.distillation.MasterConfig attribute) (dockyard_rl.algorithms.dpo.MasterConfig attribute) (dockyard_rl.algorithms.grpo.MasterConfig attribute) (dockyard_rl.algorithms.rm.MasterConfig attribute) (dockyard_rl.algorithms.sft.MasterConfig attribute) data_config (dockyard_rl.data.datasets.raw_dataset.RawDataset attribute) data_parallel_replicate_size (dockyard_rl.models.jax.config.JaxPolicyConfig attribute) data_parallel_size (dockyard_rl.cluster.placement.ParallelismSpec attribute) data_path (dockyard_rl.data.PreferenceDatasetConfig attribute) (dockyard_rl.data.ResponseDatasetConfig attribute) data_plane (dockyard_rl.algorithms.grpo.MasterConfig attribute) data_points (dockyard_rl.utils.memory_tracker.MemoryTracker attribute) data_sharding() (in module dockyard_rl.models.jax.sharding) DataConfig (class in dockyard_rl.data) DataPlaneClient (class in dockyard_rl.data_plane.interfaces) DataPlaneConfig (class in dockyard_rl.data_plane.interfaces) DataPlaneEvent (class in dockyard_rl.data_plane.observability) DataPlaneStats (class in dockyard_rl.data_plane.observability) dataset (dockyard_rl.data.datasets.raw_dataset.RawDataset attribute) dataset_name (dockyard_rl.data.AIMEEvalDataConfig attribute) (dockyard_rl.data.GPQAEvalDataConfig attribute) (dockyard_rl.data.LocalMathEvalDataConfig attribute) (dockyard_rl.data.MathEvalDataConfig attribute) (dockyard_rl.data.MMAUEvalDataConfig attribute) (dockyard_rl.data.MMMLUEvalDataConfig attribute) (dockyard_rl.data.MMMLUProEvalDataConfig attribute) (dockyard_rl.data.PreferenceDatasetConfig attribute) (dockyard_rl.data.ResponseDatasetConfig attribute) (dockyard_rl.environments.retriever.RAGEnvConfig attribute) DATASET_REGISTRY (in module dockyard_rl.data.datasets.response_datasets) dataset_split (dockyard_rl.environments.retriever.RAGEnvConfig attribute) dataset_subset (dockyard_rl.environments.retriever.RAGEnvConfig attribute) DatumSpec (class in dockyard_rl.data.interfaces) day_of_week_mapping (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.misc) decode_log_interval (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) decode_skip_special_tokens (dockyard_rl.tool_protocol.protocol.StructuredToolUseConfig attribute) decoder_sparse_step (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) (dockyard_rl.models.jax.moe.qwen3_moe.Qwen3MoeConfig attribute) decompose_message_log() (in module dockyard_rl.data.llm_message_utils) deduplicate_shared_teacher_checkpoints (dockyard_rl.algorithms.opd.OnPolicyDistillationConfig attribute) DeepScalerDataset (class in dockyard_rl.data.datasets.response_datasets.deepscaler) default (dockyard_rl.data.DataConfig attribute) DEFAULT_ABSTENTION_TEXT (in module dockyard_rl.algorithms.rlaif) DEFAULT_CHAR_LIMIT (in module dockyard_rl.environments._gdpval_extract) DEFAULT_CONSTITUTION (in module dockyard_rl.algorithms.rlaif) DEFAULT_FIXED_OVERHEAD_BYTES (in module dockyard_rl.algorithms.distillation_memory) DEFAULT_JWT_AUDIENCE (in module dockyard_rl.experience.cua.clients.control_center) DEFAULT_JWT_ISSUER (in module dockyard_rl.experience.cua.clients.control_center) DEFAULT_MASTER_PORT_RANGE_HIGH (in module dockyard_rl.distributed.virtual_cluster) DEFAULT_MASTER_PORT_RANGE_LOW (in module dockyard_rl.distributed.virtual_cluster) DEFAULT_MAX_IMAGE_PIXELS (in module dockyard_rl.data.multimodal_utils) DEFAULT_MEDIA_EXTENSIONS (in module dockyard_rl.data.multimodal_utils) DEFAULT_NVFP4_IGNORE (in module dockyard_rl.modelopt.utils) DEFAULT_OPTIMIZER_BYTES_PER_PARAM (in module dockyard_rl.algorithms.distillation_memory) DEFAULT_PORT (in module dockyard_rl.experience.cua.clients.control_center) default_processor (dockyard_rl.environments.utils.EnvRegistryEntry attribute) DEFAULT_SERVER_PORT (in module dockyard_rl.experience.cua.clients.guest_server) default_teacher_alias (dockyard_rl.algorithms.opd.OnPolicyDistillationConfig attribute) default_teacher_cfg (dockyard_rl.algorithms.opd.NonColocatedTeachersConfig attribute) DEFAULT_TEMPLATE (in module dockyard_rl.data.datasets.response_datasets.avqa) DEFAULT_THINKING_TAGS (in module dockyard_rl.rewards.invalid_action) DEFAULT_TOKEN_DISPATCHER (in module dockyard_rl.models.dtensor.moe.config) DEFAULT_TRAINER_ENV (in module dockyard_rl.models.jax.config) define_metric() (dockyard_rl.utils.logger.WandbLogger method) delete_session() (in module dockyard_rl.sandbox.task_executor_client) delete_task() (in module dockyard_rl.sandbox.task_executor_client) deliverable_dir (dockyard_rl.environments.gdpval_agentic_environment.GDPvalAgenticEnvConfig attribute) dense_cfg (dockyard_rl.models.jax.moe.qwen3_moe.Qwen3MoeConfig property) derive_structural_submask() (in module dockyard_rl.models.generation.vllm.utils) describe() (dockyard_rl.utils.device_capability.DeviceCapability method) deserialize() (dockyard_rl.models.generation.sglang.sglang_copied_utils.MultiprocessingSerializer static method) desirable_weight (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.loss.loss_functions.DPOLossConfig attribute) destroy() (dockyard_rl.distributed.stateless_process_group.StatelessProcessGroup method) detail (dockyard_rl.rewards.invalid_action.Violation attribute) detect_invalid_tool_call() (in module dockyard_rl.rewards.invalid_action) detect_malformed_thinking() (in module dockyard_rl.rewards.invalid_action) device_id_to_physical_device_id() (in module dockyard_rl.utils.device_query) device_uuid() (dockyard_rl.utils.device_query.DeviceQueryBackend method) (dockyard_rl.utils.device_query.NvmlBackend method) DeviceCapability (class in dockyard_rl.utils.device_capability) DeviceQueryBackend (class in dockyard_rl.utils.device_query) diff_text_file() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) difficulty_exponent (dockyard_rl.algorithms.effort_reward.EffortRewardConfig attribute) difficulty_floor (dockyard_rl.algorithms.effort_reward.EffortRewardConfig attribute) disable_cuda_graph (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) disable_cuda_graph_padding (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) disable_custom_all_reduce (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) disable_outlines_disk_cache (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) disable_overlap_schedule (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) disable_piecewise_cuda_graph (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) disable_ppo_ratio (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) disable_radix_cache (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) disable_student_only_features() (in module dockyard_rl.models.policy.teacher_worker_group) discard_samples() (dockyard_rl.models.policy.tq_policy.TQPolicy method) discover_roce_devices() (in module dockyard_rl.data_plane.rdma) dispatch() (dockyard_rl.models.dtensor.moe.dispatch.AllToAllTokenDispatcher method) (dockyard_rl.models.dtensor.moe.dispatch.LocalTokenDispatcher method) (dockyard_rl.models.jax.moe.alltoall.AllToAllTokenDispatcher method) (dockyard_rl.models.jax.moe.dispatch.LocalTokenDispatcher method) distillation (dockyard_rl.algorithms.distillation.MasterConfig attribute) DISTILLATION (dockyard_rl.algorithms.loss.interfaces.LossInputType attribute) DISTILLATION_CROSS_TOKENIZER (dockyard_rl.algorithms.loss.interfaces.LossInputType attribute) distillation_train() (in module dockyard_rl.algorithms.distillation) DistillationConfig (class in dockyard_rl.algorithms.distillation) DistillationLossConfig (class in dockyard_rl.algorithms.loss.loss_functions) DistillationLossDataDict (class in dockyard_rl.algorithms.loss.loss_functions) DistillationLossFn (class in dockyard_rl.algorithms.loss.loss_functions) DistillationSaveState (class in dockyard_rl.algorithms.distillation) distributed_vocab_topk() (in module dockyard_rl.distributed.model_utils) DistributedCrossEntropy (class in dockyard_rl.distributed.model_utils) DistributedLogprob (class in dockyard_rl.distributed.model_utils) DistributedLogprobWithSampling (class in dockyard_rl.distributed.model_utils) dockerfile (dockyard_rl.sandbox.task_executor_client.SessionStartSpec attribute) DockerPoolProvisioner (class in dockyard_rl.experience.cua.provisioner) DOCKYARD_NSYS_PROFILE_STEP_RANGE (in module dockyard_rl.distributed.worker_groups_utils) (in module dockyard_rl.utils.nsys) DOCKYARD_NSYS_WORKER_PATTERNS (in module dockyard_rl.distributed.worker_groups_utils) (in module dockyard_rl.utils.nsys) dockyard_rl module dockyard_rl.algorithms module dockyard_rl.algorithms.advantage_estimator module dockyard_rl.algorithms.async_utils module dockyard_rl.algorithms.async_utils.interfaces module dockyard_rl.algorithms.async_utils.replay_buffer module dockyard_rl.algorithms.async_utils.trajectory_collector module dockyard_rl.algorithms.config_access module dockyard_rl.algorithms.distillation module dockyard_rl.algorithms.distillation_memory module dockyard_rl.algorithms.dpo module dockyard_rl.algorithms.dpo_evals module dockyard_rl.algorithms.effort_reward module dockyard_rl.algorithms.grpo module dockyard_rl.algorithms.logits_sampling_utils module dockyard_rl.algorithms.loss module dockyard_rl.algorithms.loss.interfaces module dockyard_rl.algorithms.loss.loss_functions module dockyard_rl.algorithms.loss.utils module dockyard_rl.algorithms.loss.wrapper module dockyard_rl.algorithms.online_dpo module dockyard_rl.algorithms.opd module dockyard_rl.algorithms.reward_functions module dockyard_rl.algorithms.rlaif module dockyard_rl.algorithms.rm module dockyard_rl.algorithms.rollout_allocation module dockyard_rl.algorithms.routing_diversity module dockyard_rl.algorithms.sft module dockyard_rl.algorithms.teacher_panel module dockyard_rl.algorithms.utils module dockyard_rl.algorithms.x_token module dockyard_rl.algorithms.x_token.loss_utils module dockyard_rl.algorithms.x_token.token_aligner module dockyard_rl.algorithms.x_token.utils module dockyard_rl.cluster module dockyard_rl.cluster.bootstrap module dockyard_rl.cluster.fleet module dockyard_rl.cluster.placement module dockyard_rl.data module dockyard_rl.data.chat_templates module dockyard_rl.data.collate_fn module dockyard_rl.data.cross_tokenizer_collate module dockyard_rl.data.dataloader module dockyard_rl.data.datasets module dockyard_rl.data.datasets.eval_datasets module dockyard_rl.data.datasets.eval_datasets.aime module dockyard_rl.data.datasets.eval_datasets.gpqa module dockyard_rl.data.datasets.eval_datasets.local_math_dataset module dockyard_rl.data.datasets.eval_datasets.math module dockyard_rl.data.datasets.eval_datasets.mmau module dockyard_rl.data.datasets.eval_datasets.mmlu module dockyard_rl.data.datasets.eval_datasets.mmlu_pro module dockyard_rl.data.datasets.gdpval module dockyard_rl.data.datasets.hle module dockyard_rl.data.datasets.preference_datasets module dockyard_rl.data.datasets.preference_datasets.binary_preference_dataset module dockyard_rl.data.datasets.preference_datasets.helpsteer3 module dockyard_rl.data.datasets.preference_datasets.preference_dataset module dockyard_rl.data.datasets.preference_datasets.tulu3 module dockyard_rl.data.datasets.processed_dataset module dockyard_rl.data.datasets.program_bench module dockyard_rl.data.datasets.raw_dataset module dockyard_rl.data.datasets.response_datasets module dockyard_rl.data.datasets.response_datasets.aime24 module dockyard_rl.data.datasets.response_datasets.avqa module dockyard_rl.data.datasets.response_datasets.clevr module dockyard_rl.data.datasets.response_datasets.daily_omni module dockyard_rl.data.datasets.response_datasets.dapo_math module dockyard_rl.data.datasets.response_datasets.deepscaler module dockyard_rl.data.datasets.response_datasets.gemometry3k module dockyard_rl.data.datasets.response_datasets.general_conversation_dataset module dockyard_rl.data.datasets.response_datasets.gsm8k module dockyard_rl.data.datasets.response_datasets.helpsteer3 module dockyard_rl.data.datasets.response_datasets.oai_format_dataset module dockyard_rl.data.datasets.response_datasets.oasst module dockyard_rl.data.datasets.response_datasets.openmathinstruct2 module dockyard_rl.data.datasets.response_datasets.refcoco module dockyard_rl.data.datasets.response_datasets.response_dataset module dockyard_rl.data.datasets.response_datasets.squad module dockyard_rl.data.datasets.response_datasets.tulu3 module dockyard_rl.data.datasets.swe_bench module dockyard_rl.data.datasets.swe_bench_pro module dockyard_rl.data.datasets.terminal_bench module dockyard_rl.data.datasets.utils module dockyard_rl.data.interfaces module dockyard_rl.data.llm_message_utils module dockyard_rl.data.multimodal_utils module dockyard_rl.data.packing module dockyard_rl.data.packing.algorithms module dockyard_rl.data.packing.metrics module dockyard_rl.data.packing.packing module dockyard_rl.data.processors module dockyard_rl.data.utils module dockyard_rl.data_plane module dockyard_rl.data_plane.adapters module dockyard_rl.data_plane.adapters.noop module dockyard_rl.data_plane.adapters.transfer_queue module dockyard_rl.data_plane.codec module dockyard_rl.data_plane.column_io module dockyard_rl.data_plane.factory module dockyard_rl.data_plane.interfaces module dockyard_rl.data_plane.observability module dockyard_rl.data_plane.preshard module dockyard_rl.data_plane.rdma module dockyard_rl.data_plane.schema module dockyard_rl.data_plane.worker_mixin module dockyard_rl.distributed module dockyard_rl.distributed.batched_data_dict module dockyard_rl.distributed.model_utils module dockyard_rl.distributed.named_sharding module dockyard_rl.distributed.stateless_process_group module dockyard_rl.distributed.virtual_cluster module dockyard_rl.distributed.worker_groups module dockyard_rl.distributed.worker_groups_utils module dockyard_rl.environments module dockyard_rl.environments._gdpval_extract module dockyard_rl.environments.code_environment module dockyard_rl.environments.code_jaccard_environment module dockyard_rl.environments.dapo_math_verifier module dockyard_rl.environments.gdpval_agentic_environment module dockyard_rl.environments.gdpval_environment module dockyard_rl.environments.hle_environment module dockyard_rl.environments.interfaces module dockyard_rl.environments.math_environment module dockyard_rl.environments.metrics module dockyard_rl.environments.multi_turn_session_env module dockyard_rl.environments.program_bench_environment module dockyard_rl.environments.retriever module dockyard_rl.environments.reward_model_environment module dockyard_rl.environments.rewards module dockyard_rl.environments.terminal_bench_environment module dockyard_rl.environments.utils module dockyard_rl.environments.vlm_environment module

dockyard_rl.evals module dockyard_rl.evals.answer_parsing module dockyard_rl.experience module dockyard_rl.experience.cua module dockyard_rl.experience.cua.actions module dockyard_rl.experience.cua.actions.cc_transpile module dockyard_rl.experience.cua.actions.pyautogui_parse module dockyard_rl.experience.cua.backends module dockyard_rl.experience.cua.backends.control_center module dockyard_rl.experience.cua.backends.interface module dockyard_rl.experience.cua.backends.macos module dockyard_rl.experience.cua.backends.official module dockyard_rl.experience.cua.backends.windows module dockyard_rl.experience.cua.clients module dockyard_rl.experience.cua.clients.control_center module dockyard_rl.experience.cua.clients.guest_server module dockyard_rl.experience.cua.clients.proto module dockyard_rl.experience.cua.datasets module dockyard_rl.experience.cua.datasets.osworld module dockyard_rl.experience.cua.environment module dockyard_rl.experience.cua.grading module dockyard_rl.experience.cua.grading.builtins module dockyard_rl.experience.cua.grading.dispatch module dockyard_rl.experience.cua.grading.factory module dockyard_rl.experience.cua.grading.registry module dockyard_rl.experience.cua.provisioner module dockyard_rl.experience.cua.rollout module dockyard_rl.experience.cua.tasks module dockyard_rl.experience.cua.tasks.validate module dockyard_rl.experience.cua.vendor module dockyard_rl.experience.cua.vendor.desktop_env_eval module dockyard_rl.experience.cua.vendor.desktop_env_eval.dispatch module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.file module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.general module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.info module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.misc module dockyard_rl.experience.cua.vendor.desktop_env_eval.grading_env module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.basic_os module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.pdf module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.table module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.utils module dockyard_rl.experience.cua.vision module dockyard_rl.experience.rollouts module dockyard_rl.modelopt module dockyard_rl.modelopt.registry module dockyard_rl.modelopt.utils module dockyard_rl.models module dockyard_rl.models.dtensor module dockyard_rl.models.dtensor.moe module dockyard_rl.models.dtensor.moe.block module dockyard_rl.models.dtensor.moe.config module dockyard_rl.models.dtensor.moe.convert module dockyard_rl.models.dtensor.moe.detect module dockyard_rl.models.dtensor.moe.dispatch module dockyard_rl.models.dtensor.moe.experts module dockyard_rl.models.dtensor.moe.load_balance module dockyard_rl.models.dtensor.moe.mesh module dockyard_rl.models.dtensor.moe.parallelize_moe module dockyard_rl.models.dtensor.moe.refit module dockyard_rl.models.dtensor.moe.router module dockyard_rl.models.dtensor.moe.router_replay module dockyard_rl.models.dtensor.moe.sharding module dockyard_rl.models.dtensor.moe.surgery module dockyard_rl.models.dtensor.parallelize module dockyard_rl.models.generation module dockyard_rl.models.generation.constants module dockyard_rl.models.generation.interfaces module dockyard_rl.models.generation.sglang module dockyard_rl.models.generation.sglang.config module dockyard_rl.models.generation.sglang.sglang_copied_utils module dockyard_rl.models.generation.sglang.sglang_generation module dockyard_rl.models.generation.sglang.sglang_worker module dockyard_rl.models.generation.sglang.utils module dockyard_rl.models.generation.vllm module dockyard_rl.models.generation.vllm.config module dockyard_rl.models.generation.vllm.quantization module dockyard_rl.models.generation.vllm.quantization.fp8 module dockyard_rl.models.generation.vllm.quantization.fp8_train_utils module dockyard_rl.models.generation.vllm.router_capture module dockyard_rl.models.generation.vllm.utils module dockyard_rl.models.generation.vllm.vllm_backend module dockyard_rl.models.generation.vllm.vllm_generation module dockyard_rl.models.generation.vllm.vllm_worker module dockyard_rl.models.generation.vllm.vllm_worker_async module dockyard_rl.models.huggingface module dockyard_rl.models.huggingface.common module dockyard_rl.models.jax module dockyard_rl.models.jax.checkpoint module dockyard_rl.models.jax.config module dockyard_rl.models.jax.convert module dockyard_rl.models.jax.layers module dockyard_rl.models.jax.linear_attn module dockyard_rl.models.jax.linear_attn.delta_rule module dockyard_rl.models.jax.linear_attn.gated_deltanet module dockyard_rl.models.jax.linear_attn.refit module dockyard_rl.models.jax.linear_attn.weights module dockyard_rl.models.jax.loss module dockyard_rl.models.jax.models module dockyard_rl.models.jax.models.gemma4 module dockyard_rl.models.jax.models.gemma4_refit module dockyard_rl.models.jax.models.qwen3 module dockyard_rl.models.jax.models.qwen3_next module dockyard_rl.models.jax.moe module dockyard_rl.models.jax.moe.alltoall module dockyard_rl.models.jax.moe.block module dockyard_rl.models.jax.moe.dispatch module dockyard_rl.models.jax.moe.experts module dockyard_rl.models.jax.moe.load_balance module dockyard_rl.models.jax.moe.qwen3_moe module dockyard_rl.models.jax.moe.refit module dockyard_rl.models.jax.moe.router module dockyard_rl.models.jax.moe.weights module dockyard_rl.models.jax.optimizer module dockyard_rl.models.jax.policy_worker module dockyard_rl.models.jax.refit module dockyard_rl.models.jax.sharding module dockyard_rl.models.jax.train_step module dockyard_rl.models.jax.weights module dockyard_rl.models.policy module dockyard_rl.models.policy.interfaces module dockyard_rl.models.policy.lm_policy module dockyard_rl.models.policy.teacher_scorer module dockyard_rl.models.policy.teacher_worker_group module dockyard_rl.models.policy.tq_policy module dockyard_rl.models.policy.utils module dockyard_rl.models.policy.workers module dockyard_rl.models.policy.workers.base_policy_worker module dockyard_rl.models.policy.workers.dtensor_policy_worker module dockyard_rl.models.policy.workers.dtensor_policy_worker_v2 module dockyard_rl.rewards module dockyard_rl.rewards.compiler module dockyard_rl.rewards.diff_utils module dockyard_rl.rewards.gdpval_rubric module dockyard_rl.rewards.hle_grader module dockyard_rl.rewards.integrity module dockyard_rl.rewards.interfaces module dockyard_rl.rewards.invalid_action module dockyard_rl.rewards.program_bench module dockyard_rl.rewards.swe_bench_pro module dockyard_rl.rewards.terminal_bench module dockyard_rl.rewards.test_runner module dockyard_rl.sandbox module dockyard_rl.sandbox.task_executor_client module dockyard_rl.tool_protocol module dockyard_rl.tool_protocol.hermes module dockyard_rl.tool_protocol.protocol module dockyard_rl.tool_protocol.registry module dockyard_rl.tool_protocol.schema module dockyard_rl.utils module dockyard_rl.utils.checkpoint module dockyard_rl.utils.config module dockyard_rl.utils.device_capability module dockyard_rl.utils.device_query module dockyard_rl.utils.gguf_export module dockyard_rl.utils.logger module dockyard_rl.utils.memory_tracker module dockyard_rl.utils.native_checkpoint module dockyard_rl.utils.nsys module dockyard_rl.utils.packed_tensor module dockyard_rl.utils.timer module dockyard_rl.weight_sync module dockyard_rl.weight_sync.collective_weight_synchronizer module dockyard_rl.weight_sync.factory module dockyard_rl.weight_sync.http_weight_synchronizer module dockyard_rl.weight_sync.interfaces module dockyard_rl.weight_sync.ipc_weight_synchronizer module download_dir (dockyard_rl.data.ResponseDatasetConfig attribute) DP dp (dockyard_rl.models.dtensor.moe.mesh.MoEParallelDims property) (dockyard_rl.models.jax.sharding.MoEMeshDims attribute) DP_CALIB_INPUT_FIELDS (in module dockyard_rl.data_plane.schema) dp_packing() (in module dockyard_rl.data.packing.algorithms) dp_replicate (dockyard_rl.models.dtensor.moe.mesh.MoEParallelDims attribute) dp_shard (dockyard_rl.models.dtensor.moe.mesh.MoEParallelDims attribute) dp_size (dockyard_rl.distributed.worker_groups.RayWorkerGroup property) (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) DP_TRAIN_FIELDS (in module dockyard_rl.data_plane.schema) DPO dpo (dockyard_rl.algorithms.dpo.MasterConfig attribute) dpo_train() (in module dockyard_rl.algorithms.dpo) DPOConfig (class in dockyard_rl.algorithms.dpo) DPOLossConfig (class in dockyard_rl.algorithms.loss.loss_functions) DPOLossDataDict (class in dockyard_rl.algorithms.loss.loss_functions) DPOLossFn (class in dockyard_rl.algorithms.loss.loss_functions) DPOP dpop_lambda (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.loss.loss_functions.DPOLossConfig attribute) DPOPLossFn (class in dockyard_rl.algorithms.loss.loss_functions) DPOSaveState (class in dockyard_rl.algorithms.dpo) DPOValMetrics (class in dockyard_rl.algorithms.dpo) DRA DRAFT (dockyard_rl.algorithms.loss.interfaces.LossInputType attribute) DraftCrossEntropyLossConfig (class in dockyard_rl.algorithms.loss.loss_functions) DraftCrossEntropyLossDataDict (class in dockyard_rl.algorithms.loss.loss_functions) DraftCrossEntropyLossFn (class in dockyard_rl.algorithms.loss.loss_functions) DraftLossWrapper (class in dockyard_rl.algorithms.loss.wrapper) drop_ray_visible_device_opt_outs() (in module dockyard_rl.utils.device_query) DTensor dtensor_cfg (dockyard_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) dtensor_cfg_overrides (dockyard_rl.algorithms.opd.TeacherResourceConfig attribute) (dockyard_rl.models.policy.teacher_worker_group.TeacherConfig attribute) dtensor_from_parallel_logits_to_logprobs() (in module dockyard_rl.distributed.model_utils) DTensorPolicyWorker (class in dockyard_rl.models.policy.workers.dtensor_policy_worker) DTensorPolicyWorkerImpl (class in dockyard_rl.models.policy.workers.dtensor_policy_worker) DTensorPolicyWorkerV2 (class in dockyard_rl.models.policy.workers.dtensor_policy_worker_v2) DTensorPolicyWorkerV2Impl (class in dockyard_rl.models.policy.workers.dtensor_policy_worker_v2) dtype (dockyard_rl.algorithms.distillation.TeacherConfig attribute) (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) DTYPE_BYTES (in module dockyard_rl.algorithms.distillation_memory) dtype_bytes() (in module dockyard_rl.algorithms.distillation_memory) dynamic_batching (dockyard_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) dynamic_loss_scaling (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) dynamic_sampling() (in module dockyard_rl.algorithms.grpo) dynamic_sampling_max_gen_batches (dockyard_rl.algorithms.grpo.GRPOConfig attribute) DynamicBatchingArgs (class in dockyard_rl.distributed.batched_data_dict)

E

effective_coefficient() (in module dockyard_rl.algorithms.effort_reward) effort_reward (dockyard_rl.algorithms.grpo.GRPOConfig attribute) effort_reward_metrics() (in module dockyard_rl.algorithms.effort_reward) EffortRewardConfig (class in dockyard_rl.algorithms.effort_reward) efsdp (dockyard_rl.models.dtensor.moe.mesh.MoEParallelDims property) ELEM_COUNTS_PER_GB (in module dockyard_rl.data_plane.schema) embed_scale (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig property) Embedding (class in dockyard_rl.models.jax.layers) empty_like() (dockyard_rl.data.multimodal_utils.PackedTensor class method) enable_dp_attention (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) enable_ep (dockyard_rl.models.dtensor.moe.mesh.MoEParallelDims property) (dockyard_rl.models.jax.sharding.MoEMeshDims property) enable_ep_moe (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) enable_fast_load (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) enable_lora (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) enable_memory_saver (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) enable_metrics (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) enable_mixed_chunk (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) enable_moe_block (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) enable_multimodal (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) enable_multithread_load (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) enable_nan_detection (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) enable_nccl_nvls (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) enable_p2p_check (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) enable_prefix_caching (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) enable_schema_validation (dockyard_rl.rewards.invalid_action.InvalidActionPenaltyConfig attribute) enable_torch_compile (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) enable_vllm_metrics_logger (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) enabled (dockyard_rl.algorithms.effort_reward.EffortRewardConfig attribute) (dockyard_rl.algorithms.grpo.AsyncGRPOConfig attribute) (dockyard_rl.algorithms.grpo.RewardScalingConfig attribute) (dockyard_rl.algorithms.online_dpo.PreferenceJudge property) (dockyard_rl.algorithms.opd.NonColocatedTeachersConfig attribute) (dockyard_rl.algorithms.opd.OnPolicyDistillationConfig attribute) (dockyard_rl.algorithms.opd.TeacherPanelConfig attribute) (dockyard_rl.algorithms.reward_functions.RewardShapingConfig attribute) (dockyard_rl.algorithms.rlaif.ChatLLM attribute) (dockyard_rl.algorithms.rollout_allocation.GradedAllocationConfig attribute) (dockyard_rl.algorithms.routing_diversity.RoutingDiversityConfig attribute) (dockyard_rl.data_plane.interfaces.DataPlaneConfig attribute) (dockyard_rl.data_plane.interfaces.ObservabilityConfig attribute) (dockyard_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) (dockyard_rl.models.jax.config.JaxPolicyConfig attribute) (dockyard_rl.rewards.hle_grader.HLEJudgeClient property) (dockyard_rl.rewards.invalid_action.InvalidActionPenaltyConfig attribute) (dockyard_rl.tool_protocol.protocol.StructuredToolUseConfig attribute) (dockyard_rl.utils.checkpoint.CheckpointingConfig attribute) encode_single() (dockyard_rl.data.datasets.processed_dataset.AllTaskProcessedDataset method) enforce_eager (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) EnglishMultichoiceVerifyWorker (class in dockyard_rl.environments.math_environment) entity (dockyard_rl.utils.logger.WandbConfig attribute) entry_command (dockyard_rl.sandbox.task_executor_client.TaskSpec attribute) env (dockyard_rl.algorithms.grpo.MasterConfig attribute) (dockyard_rl.sandbox.task_executor_client.SessionStartSpec attribute) ENV_INVALID_ACTION_KEY (in module dockyard_rl.rewards.invalid_action) ENV_MALFORMED_THINKING_KEY (in module dockyard_rl.rewards.invalid_action) env_name (dockyard_rl.data.MMAUEvalDataConfig attribute) (dockyard_rl.data.ResponseDatasetConfig attribute) ENV_REGISTRY (in module dockyard_rl.environments.utils) env_vars (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) (dockyard_rl.models.jax.config.JaxPolicyConfig attribute) EnvironmentInterface (class in dockyard_rl.environments.interfaces) EnvironmentReturn (class in dockyard_rl.environments.interfaces) EnvRegistryEntry (class in dockyard_rl.environments.utils) EP ep (dockyard_rl.models.dtensor.moe.mesh.MoEParallelDims attribute) (dockyard_rl.models.jax.sharding.MoEMeshDims attribute)

ep_coord() (dockyard_rl.models.dtensor.moe.mesh.MoEParallelDims method) ep_size (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) epoch (dockyard_rl.algorithms.distillation.DistillationSaveState attribute) (dockyard_rl.algorithms.dpo.DPOSaveState attribute) (dockyard_rl.algorithms.rm.RMSaveState attribute) (dockyard_rl.algorithms.sft.SFTSaveState attribute) error (dockyard_rl.sandbox.task_executor_client.FinishResult attribute) (dockyard_rl.tool_protocol.hermes.HermesExtraction attribute) escalation_fraction (dockyard_rl.algorithms.opd.TeacherPanelConfig attribute) estimate_activation_bytes() (in module dockyard_rl.algorithms.distillation_memory) eval_collate_fn() (in module dockyard_rl.data.collate_fn) EVAL_DATASET_REGISTRY (in module dockyard_rl.data.datasets.eval_datasets) EvalDataConfigType (in module dockyard_rl.data) evaluate() (dockyard_rl.experience.cua.backends.control_center.ControlCenterBackend method) (dockyard_rl.experience.cua.backends.interface.OSWorldBackend method) (dockyard_rl.experience.cua.backends.macos.MacBackend method) (dockyard_rl.experience.cua.backends.official.OfficialDesktopEnvBackend method) (dockyard_rl.experience.cua.backends.windows.WindowsBackend method) (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.dispatch) evaluate_custom() (in module dockyard_rl.experience.cua.grading.dispatch) EVALUATOR_REGISTRY (in module dockyard_rl.experience.cua.grading.registry) EvaluatorFn (in module dockyard_rl.experience.cua.grading.registry) EventStatus (in module dockyard_rl.data_plane.observability) evict() (dockyard_rl.algorithms.async_utils.interfaces.ReplayBufferProtocol method) (dockyard_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) evidence_hash (dockyard_rl.rewards.interfaces.RewardVerificationResult attribute) exact_answer_alphanumeric_reward() (in module dockyard_rl.environments.rewards) exact_match() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) (in module dockyard_rl.rewards.hle_grader) exact_token_match_only (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) ExecResult (class in dockyard_rl.sandbox.task_executor_client) execute() (dockyard_rl.experience.cua.clients.control_center.ControlCenterClient method) execute_python_command() (dockyard_rl.experience.cua.clients.guest_server.GuestServerController method) (dockyard_rl.experience.cua.vendor.desktop_env_eval.grading_env.GuestServerController method) execution_time (dockyard_rl.sandbox.task_executor_client.FinishResult attribute) (dockyard_rl.sandbox.task_executor_client.TaskResult attribute) exit_code (dockyard_rl.sandbox.task_executor_client.ExecResult attribute) (dockyard_rl.sandbox.task_executor_client.FinishResult attribute) (dockyard_rl.sandbox.task_executor_client.TaskResult attribute) expand_grouped_expert_info() (in module dockyard_rl.models.dtensor.moe.refit) (in module dockyard_rl.models.jax.moe.refit) expand_grouped_expert_key() (in module dockyard_rl.models.dtensor.moe.refit) (in module dockyard_rl.models.jax.moe.refit) expected_calibration_error() (in module dockyard_rl.algorithms.dpo_evals) experiment_name (dockyard_rl.utils.logger.MLflowConfig attribute) EXPERT_DOWN_AXES (in module dockyard_rl.models.jax.sharding) EXPERT_GATE_UP_AXES (in module dockyard_rl.models.jax.sharding) expert_major_permutation() (in module dockyard_rl.models.jax.moe.alltoall) expert_parallel_size (dockyard_rl.cluster.placement.ParallelismSpec attribute) (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) (dockyard_rl.models.jax.config.JaxPolicyConfig attribute) ExpertRole (in module dockyard_rl.models.dtensor.moe.sharding) experts_for_rank() (dockyard_rl.models.dtensor.moe.mesh.MoEParallelDims method) export_gguf_variants() (in module dockyard_rl.utils.gguf_export) export_hf_safetensors() (in module dockyard_rl.models.jax.checkpoint) expose_http_server (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) extra_env_info (dockyard_rl.data.interfaces.DatumSpec attribute) extra_info (dockyard_rl.data_plane.interfaces.KVBatchMeta attribute) extract_answer() (in module dockyard_rl.rewards.hle_grader) extract_answers_from_pdf() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.pdf) extract_dir() (in module dockyard_rl.environments._gdpval_extract) extract_file() (in module dockyard_rl.environments._gdpval_extract) extract_initial_prompt_messages() (in module dockyard_rl.algorithms.grpo) extract_necessary_env_names() (in module dockyard_rl.data.datasets.utils) extracted_answer (dockyard_rl.environments.math_environment.MathEnvironmentMetadata attribute) extraction_char_limit (dockyard_rl.environments.gdpval_agentic_environment.GDPvalAgenticEnvConfig attribute) extraction_timeout (dockyard_rl.environments.gdpval_agentic_environment.GDPvalAgenticEnvConfig attribute)

F

FACTUALITY_SCORE_TEMPLATE (in module dockyard_rl.algorithms.rlaif) failure_reason (dockyard_rl.rewards.interfaces.RewardVerificationResult attribute) FAST (dockyard_rl.algorithms.distillation_memory.ColocationSchedule attribute) FetchPolicy (in module dockyard_rl.data_plane.worker_mixin) fields (dockyard_rl.data_plane.interfaces.KVBatchMeta attribute) file_contains() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) file_format (dockyard_rl.data.LocalMathEvalDataConfig attribute) final_action_match() (in module dockyard_rl.experience.cua.grading.builtins) final_logit_softcapping (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) finalize_checkpoint() (dockyard_rl.utils.checkpoint.CheckpointManager method) finish_generation() (dockyard_rl.models.generation.interfaces.GenerationInterface method) (dockyard_rl.models.generation.sglang.sglang_generation.SGLangGeneration method) (dockyard_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.interfaces.ColocatablePolicyInterface method) (dockyard_rl.models.policy.lm_policy.Policy method) finish_step() (dockyard_rl.models.policy.tq_policy.TQPolicy method) finish_training() (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.interfaces.ColocatablePolicyInterface method) (dockyard_rl.models.policy.interfaces.PolicyInterface method) (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) FinishResult (class in dockyard_rl.sandbox.task_executor_client) first_fit_decreasing() (in module dockyard_rl.data.packing.algorithms) fix_gpt_oss_export_transpose() (in module dockyard_rl.models.generation.vllm.vllm_backend) fixed_overhead_bytes (dockyard_rl.algorithms.distillation_memory.MemoryReport attribute) FlashAttentionKwargs (class in dockyard_rl.models.huggingface.common) FlatMessagesType (in module dockyard_rl.data.interfaces) flatten_dict() (in module dockyard_rl.utils.logger) flattened_concat() (dockyard_rl.data.multimodal_utils.PackedTensor class method) FLEET_INFERENCE (in module dockyard_rl.cluster.bootstrap) FLEET_SANDBOX (in module dockyard_rl.cluster.bootstrap) FLEET_TRAINER (in module dockyard_rl.cluster.bootstrap) FleetSpec (class in dockyard_rl.cluster.fleet) flush() (dockyard_rl.utils.logger.RayGpuMonitorLogger method) flush_interval (dockyard_rl.utils.logger.GPUMonitoringConfig attribute) force_on_policy_ratio (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) format_answer_fromtags() (in module dockyard_rl.data.datasets.response_datasets.clevr) format_clevr_cogent_dataset() (in module dockyard_rl.data.datasets.response_datasets.clevr) format_data() (dockyard_rl.data.datasets.gdpval.GDPvalDataset method) (dockyard_rl.data.datasets.hle.HLEDataset method) (dockyard_rl.data.datasets.preference_datasets.binary_preference_dataset.BinaryPreferenceDataset method) (dockyard_rl.data.datasets.preference_datasets.helpsteer3.HelpSteer3PreferenceDataset method) (dockyard_rl.data.datasets.preference_datasets.preference_dataset.PreferenceDataset method) (dockyard_rl.data.datasets.preference_datasets.tulu3.Tulu3PreferenceDataset method) (dockyard_rl.data.datasets.response_datasets.aime24.AIME2024Dataset method) (dockyard_rl.data.datasets.response_datasets.avqa.AVQADataset method) (dockyard_rl.data.datasets.response_datasets.daily_omni.DailyOmniDataset method) (dockyard_rl.data.datasets.response_datasets.dapo_math.DAPOMath17KDataset method) (dockyard_rl.data.datasets.response_datasets.deepscaler.DeepScalerDataset method) (dockyard_rl.data.datasets.response_datasets.gsm8k.GSM8KDataset method) (dockyard_rl.data.datasets.response_datasets.helpsteer3.HelpSteer3Dataset method) (dockyard_rl.data.datasets.response_datasets.oai_format_dataset.OpenAIFormatDataset method) (dockyard_rl.data.datasets.response_datasets.openmathinstruct2.OpenMathInstruct2Dataset method) (dockyard_rl.data.datasets.response_datasets.response_dataset.ResponseDataset method) (dockyard_rl.data.datasets.response_datasets.squad.SquadDataset method) (dockyard_rl.data.datasets.response_datasets.tulu3.Tulu3SftMixtureDataset method) (dockyard_rl.data.datasets.swe_bench.SWEBenchDataset method) (dockyard_rl.data.datasets.swe_bench_pro.SWEBenchProDataset method)

format_geometry3k_dataset() (in module dockyard_rl.data.datasets.response_datasets.gemometry3k) (in module dockyard_rl.data.datasets.response_datasets.refcoco) format_prompt_for_vllm_generation() (in module dockyard_rl.models.generation.vllm.utils) format_reward() (in module dockyard_rl.environments.rewards) forward() (dockyard_rl.algorithms.x_token.loss_utils.Fp32SparseMM static method) (dockyard_rl.distributed.model_utils.AllGatherCPTensor static method) (dockyard_rl.distributed.model_utils.ChunkedDistributedEntropy static method) (dockyard_rl.distributed.model_utils.ChunkedDistributedGatherLogprob static method) (dockyard_rl.distributed.model_utils.ChunkedDistributedHiddenStatesToLogprobs static method) (dockyard_rl.distributed.model_utils.ChunkedDistributedLogprob static method) (dockyard_rl.distributed.model_utils.ChunkedDistributedLogprobWithSampling static method) (dockyard_rl.distributed.model_utils.DistributedCrossEntropy static method) (dockyard_rl.distributed.model_utils.DistributedLogprob static method) (dockyard_rl.distributed.model_utils.DistributedLogprobWithSampling static method) (dockyard_rl.models.dtensor.moe.block.MoEBlock method) (dockyard_rl.models.dtensor.moe.experts.GroupedExperts method) (dockyard_rl.models.dtensor.moe.router.TokenChoiceTopKRouter method) Fp32SparseMM (class in dockyard_rl.algorithms.x_token.loss_utils) FP8 FP8_BLOCK_QUANT_KWARGS (in module dockyard_rl.models.generation.vllm.quantization.fp8) fp8_e4m3_dtype() (in module dockyard_rl.utils.device_capability) fp8_param_names (dockyard_rl.models.generation.vllm.quantization.fp8.FP8State attribute) fp8_patches_applied (in module dockyard_rl.models.generation.vllm.quantization.fp8) fp8_state (in module dockyard_rl.models.generation.vllm.quantization.fp8) FP8Config (class in dockyard_rl.models.generation.vllm.quantization.fp8) FP8State (class in dockyard_rl.models.generation.vllm.quantization.fp8) free_bytes (dockyard_rl.algorithms.distillation_memory.MemoryReport attribute) from_batches() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict class method) from_dict() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict class method) from_dtensor_cfg() (dockyard_rl.models.dtensor.moe.mesh.MoEParallelDims class method) from_env() (dockyard_rl.cluster.fleet.FleetSpec class method) from_hf_config() (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig class method) (dockyard_rl.models.jax.models.qwen3.Qwen3Config class method) (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig class method) (dockyard_rl.models.jax.moe.qwen3_moe.Qwen3MoeConfig class method) from_parallel_hidden_states_to_logprobs() (in module dockyard_rl.distributed.model_utils) from_parallel_logits_to_logprobs() (in module dockyard_rl.distributed.model_utils) from_parallel_logits_to_logprobs_packed_sequences() (in module dockyard_rl.distributed.model_utils) fsdp (dockyard_rl.models.dtensor.moe.mesh.MoEParallelDims property) FSDP2 fuse_hf_expert_state_dict() (in module dockyard_rl.models.dtensor.moe.convert) futures (dockyard_rl.distributed.worker_groups.MultiWorkerFuture attribute) fuzzy_match() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) fuzzy_place_math() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general)

G

gather_logits_at_global_indices() (in module dockyard_rl.distributed.model_utils) GDPO GDPOAdvantageEstimator (class in dockyard_rl.algorithms.advantage_estimator) GDPval gdpval_agentic_data_processor() (in module dockyard_rl.data.processors) gdpval_data_processor() (in module dockyard_rl.data.processors) GDPvalAgenticDataset (class in dockyard_rl.data.datasets.gdpval) GDPvalAgenticEnvConfig (class in dockyard_rl.environments.gdpval_agentic_environment) GDPvalAgenticEnvironment (class in dockyard_rl.environments.gdpval_agentic_environment) GDPvalDataset (class in dockyard_rl.data.datasets.gdpval) GDPvalEnvConfig (class in dockyard_rl.environments.gdpval_environment) GDPvalEnvironment (class in dockyard_rl.environments.gdpval_environment) GDPvalRubricReward (class in dockyard_rl.rewards.gdpval_rubric) GEMM gemma4_inv_freq() (in module dockyard_rl.models.jax.models.gemma4) gemma4_rope_cos_sin() (in module dockyard_rl.models.jax.models.gemma4) gemma4_sliding_bias() (in module dockyard_rl.models.jax.models.gemma4) Gemma4Attention (class in dockyard_rl.models.jax.models.gemma4) Gemma4DecoderLayer (class in dockyard_rl.models.jax.models.gemma4) Gemma4Experts (class in dockyard_rl.models.jax.models.gemma4) Gemma4ForCausalLM (class in dockyard_rl.models.jax.models.gemma4) Gemma4MLP (class in dockyard_rl.models.jax.models.gemma4) Gemma4Model (class in dockyard_rl.models.jax.models.gemma4) Gemma4PerLayerEmbeddings (class in dockyard_rl.models.jax.models.gemma4) Gemma4RMSNorm (class in dockyard_rl.models.jax.models.gemma4) Gemma4Router (class in dockyard_rl.models.jax.models.gemma4) Gemma4ScaledWordEmbedding (class in dockyard_rl.models.jax.models.gemma4) Gemma4TextConfig (class in dockyard_rl.models.jax.models.gemma4) GeneralConversationsJsonlDataset (class in dockyard_rl.data.datasets.response_datasets.general_conversation_dataset) generate() (dockyard_rl.models.generation.interfaces.GenerationInterface method) (dockyard_rl.models.generation.sglang.sglang_generation.SGLangGeneration method) (dockyard_rl.models.generation.sglang.sglang_worker.SGLangGenerationWorker method) (dockyard_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (dockyard_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) (dockyard_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) (dockyard_rl.models.policy.lm_policy.Policy method) generate_async() (dockyard_rl.models.generation.sglang.sglang_generation.SGLangGeneration method) (dockyard_rl.models.generation.sglang.sglang_worker.SGLangGenerationWorker method) (dockyard_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (dockyard_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) generate_constitutional_revision() (in module dockyard_rl.algorithms.rlaif) generate_responses() (in module dockyard_rl.experience.rollouts) generate_responses_async() (in module dockyard_rl.experience.rollouts) generate_text() (dockyard_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) GenerateFn (in module dockyard_rl.algorithms.online_dpo) generation (dockyard_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) generation_lengths (dockyard_rl.models.generation.interfaces.GenerationOutputSpec attribute) generation_logprobs (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossDataDict attribute) GENERATION_WORKER_OVERRIDES (in module dockyard_rl.models.generation.vllm.utils) GenerationConfig (class in dockyard_rl.models.generation.interfaces) GenerationDatumSpec (class in dockyard_rl.models.generation.interfaces) GenerationInterface (class in dockyard_rl.models.generation.interfaces) GenerationOutputSpec (class in dockyard_rl.models.generation.interfaces) Geometry3KDataset (class in dockyard_rl.data.datasets.response_datasets.gemometry3k) (class in dockyard_rl.data.datasets.response_datasets.refcoco) get() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) (dockyard_rl.models.policy.interfaces.Timer method) (dockyard_rl.tool_protocol.registry.ToolRegistry method) get_accessibility_tree() (dockyard_rl.experience.cua.clients.guest_server.GuestServerController method) (dockyard_rl.experience.cua.vendor.desktop_env_eval.grading_env.GuestServerController method) (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.misc) get_all_worker_results() (dockyard_rl.distributed.worker_groups.RayWorkerGroup method) get_available_address_and_port() (dockyard_rl.distributed.virtual_cluster.RayVirtualCluster method) get_axis_index() (dockyard_rl.distributed.named_sharding.NamedSharding method) get_axis_size() (dockyard_rl.distributed.named_sharding.NamedSharding method) get_base_url() (dockyard_rl.models.generation.sglang.sglang_worker.SGLangGenerationWorker method) get_batch() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) get_best_checkpoint_path() (dockyard_rl.utils.checkpoint.CheckpointManager method) get_cache_file() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.file) get_cloud_file() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.file) get_content_from_vm_file() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.file) get_cpu_state_dict() (in module dockyard_rl.models.policy.workers.dtensor_policy_worker) get_data() (dockyard_rl.data_plane.adapters.noop.NoOpDataPlaneClient method) (dockyard_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) (dockyard_rl.data_plane.interfaces.DataPlaneClient method) (dockyard_rl.data_plane.observability.MetricsDataPlaneClient method) get_data_records() (in module dockyard_rl.data.datasets.response_datasets.oasst) get_dataloader_state() (dockyard_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollectorImpl method) get_debug_info() (dockyard_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) get_device_uuid() (in module dockyard_rl.utils.device_query) get_dim_to_pack_along() (in module dockyard_rl.data.multimodal_utils) get_distillation_topk_logprobs_from_logits() (in module dockyard_rl.distributed.model_utils) get_dp_leader_worker_idx() (dockyard_rl.distributed.worker_groups.RayWorkerGroup method) get_elapsed() (dockyard_rl.utils.timer.Timer method) get_existing_target_weights() (dockyard_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) get_file() (dockyard_rl.experience.cua.clients.guest_server.GuestServerController method) (dockyard_rl.experience.cua.vendor.desktop_env_eval.grading_env.GuestServerController method) get_first_index_that_differs() (in module dockyard_rl.data.llm_message_utils) get_flash_attention_kwargs() (in module dockyard_rl.models.huggingface.common) get_fleet_role() (in module dockyard_rl.cluster.bootstrap) get_formatted_message_log() (in module dockyard_rl.data.llm_message_utils) get_free_memory_bytes() (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) (in module dockyard_rl.utils.device_query) get_gdpo_reward_component_keys() (in module dockyard_rl.algorithms.utils) get_gpu_id() (dockyard_rl.distributed.virtual_cluster.GetGPUIdActor method) get_gpu_info() (dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) (in module dockyard_rl.models.policy.utils) get_gpu_uuids() (dockyard_rl.models.generation.sglang.sglang_worker.SGLangGenerationWorker method) get_grad_norm() (in module dockyard_rl.models.dtensor.parallelize) get_handle_from_tensor() (in module dockyard_rl.models.policy.utils) get_hf_tp_plan() (in module dockyard_rl.models.dtensor.parallelize) get_huggingface_cache_path() (in module dockyard_rl.data.datasets.utils) get_ipc_zmq_address() (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) get_keys_from_message_log() (in module dockyard_rl.data.llm_message_utils) get_last_target_weight_already_generated() (dockyard_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) get_latest_checkpoint_path() (dockyard_rl.utils.checkpoint.CheckpointManager method) get_latest_elapsed() (dockyard_rl.utils.timer.Timer method) get_list_directory() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.info) get_logger_metrics() (dockyard_rl.models.generation.interfaces.GenerationInterface method) (dockyard_rl.models.generation.vllm.vllm_generation.VllmGeneration method) get_logprobs() (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.interfaces.PolicyInterface method) (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.teacher_worker_group.TeacherWorkerGroup method) (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) get_logprobs_from_meta() (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.tq_policy.TQPolicy method) get_logprobs_from_vocab_parallel_logits() (in module dockyard_rl.distributed.model_utils) get_logprobs_presharded() (dockyard_rl.data_plane.worker_mixin.TQWorkerMixin method) get_master_address_and_port() (dockyard_rl.distributed.virtual_cluster.RayVirtualCluster method) get_media_from_message() (in module dockyard_rl.data.multimodal_utils) get_microbatch_iterator_dynamic_shapes_len() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) get_microbatch_iterator_for_packable_sequences_len() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) get_model_parameter_count() (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) get_multimodal_default_settings_from_processor() (in module dockyard_rl.data.multimodal_utils) get_multimodal_dict() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) get_multimodal_keys_from_processor() (in module dockyard_rl.data.multimodal_utils) get_next_experiment_dir() (in module dockyard_rl.utils.logger) get_next_token_logprobs_from_logits() (in module dockyard_rl.distributed.model_utils) get_nsight_config_if_pattern_matches() (in module dockyard_rl.distributed.worker_groups_utils) get_num_buffers() (in module dockyard_rl.utils.packed_tensor)

get_packable_sequence_order() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) get_placement_groups() (dockyard_rl.distributed.virtual_cluster.RayVirtualCluster method) get_pretrained_paths() (dockyard_rl.utils.checkpoint.CheckpointManager static method) get_prompt() (dockyard_rl.data.datasets.response_datasets.daily_omni.DailyOmniDataset class method) get_ranks() (dockyard_rl.distributed.named_sharding.NamedSharding method) get_ranks_by_coord() (dockyard_rl.distributed.named_sharding.NamedSharding method) get_ray_cluster_topology() (in module dockyard_rl.distributed.virtual_cluster) get_reference_policy_logprobs() (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.interfaces.PolicyInterface method) (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) get_reference_policy_logprobs_from_meta() (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.tq_policy.TQPolicy method) get_reference_policy_logprobs_presharded() (dockyard_rl.data_plane.worker_mixin.TQWorkerMixin method) get_result() (in module dockyard_rl.sandbox.task_executor_client) get_results() (dockyard_rl.distributed.worker_groups.MultiWorkerFuture method) get_resume_paths() (dockyard_rl.utils.checkpoint.CheckpointManager static method) get_rule() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.misc) get_rule_relativeTime() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.misc) get_runtime_env_for_policy_worker() (in module dockyard_rl.models.policy.utils) get_samples() (dockyard_rl.data_plane.adapters.noop.NoOpDataPlaneClient method) (dockyard_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) (dockyard_rl.data_plane.interfaces.DataPlaneClient method) (dockyard_rl.data_plane.observability.MetricsDataPlaneClient method) get_session_status() (in module dockyard_rl.sandbox.task_executor_client) get_sglang_server_urls() (dockyard_rl.models.generation.sglang.sglang_generation.SGLangGeneration method) get_sglang_url_to_gpu_uuids() (dockyard_rl.models.generation.sglang.sglang_generation.SGLangGeneration method) get_snapshot_str() (dockyard_rl.utils.memory_tracker.MemoryTrackerDataPoint method) get_sparse_projection_matrix() (in module dockyard_rl.algorithms.x_token.loss_utils) get_target_packed_tensor_size() (in module dockyard_rl.utils.packed_tensor) get_teacher_routing_metrics() (in module dockyard_rl.algorithms.opd) get_terminal_output() (dockyard_rl.experience.cua.clients.guest_server.GuestServerController method) (dockyard_rl.experience.cua.vendor.desktop_env_eval.grading_env.GuestServerController method) get_time_diff_range() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.misc) get_timezone_from_config() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.misc) get_timezone_from_ip() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.misc) get_timing_metrics() (dockyard_rl.utils.timer.Timer method) get_tokenizer() (in module dockyard_rl.algorithms.utils) get_topk_logits() (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.interfaces.PolicyInterface method) (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) get_topk_projection() (in module dockyard_rl.algorithms.x_token.loss_utils) get_total_memory_bytes() (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) (in module dockyard_rl.utils.device_query) get_vllm_qkv_scale_names() (in module dockyard_rl.models.generation.vllm.quantization.fp8_train_utils) get_vm_command_error() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.general) get_vm_command_line() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.general) get_vm_directory_tree() (dockyard_rl.experience.cua.clients.guest_server.GuestServerController method) (dockyard_rl.experience.cua.vendor.desktop_env_eval.grading_env.GuestServerController method) get_vm_file() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.file) get_vm_screen_size() (dockyard_rl.experience.cua.clients.guest_server.GuestServerController method) (dockyard_rl.experience.cua.vendor.desktop_env_eval.grading_env.GuestServerController method) (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.info) get_vm_terminal_output() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.general) get_vm_wallpaper() (dockyard_rl.experience.cua.clients.guest_server.GuestServerController method) (dockyard_rl.experience.cua.vendor.desktop_env_eval.grading_env.GuestServerController method) (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.info) get_vm_window_size() (dockyard_rl.experience.cua.clients.guest_server.GuestServerController method) (dockyard_rl.experience.cua.vendor.desktop_env_eval.grading_env.GuestServerController method) (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.info) get_worker_coords() (dockyard_rl.distributed.named_sharding.NamedSharding method) get_zmq_address() (dockyard_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) GetGPUIdActor (class in dockyard_rl.distributed.virtual_cluster) GLOBAL_FORWARD_PAD_SEQLEN (in module dockyard_rl.data_plane.schema) global_fp8_config (in module dockyard_rl.models.generation.vllm.quantization.fp8) global_head_dim (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) global_partial_rotary_factor (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) global_post_process_and_metrics() (dockyard_rl.environments.code_environment.CodeEnvironment method) (dockyard_rl.environments.code_jaccard_environment.CodeJaccardEnvironment method) (dockyard_rl.environments.gdpval_environment.GDPvalEnvironment method) (dockyard_rl.environments.hle_environment.HLEEnvironment method) (dockyard_rl.environments.interfaces.EnvironmentInterface method) (dockyard_rl.environments.math_environment.BaseMathEnvironment method) (dockyard_rl.environments.multi_turn_session_env.MultiTurnSessionEnvironment method) (dockyard_rl.environments.retriever.RAGEnvironment method) (dockyard_rl.environments.reward_model_environment.RewardModelEnvironment method) (dockyard_rl.environments.vlm_environment.VLMEnvironment method) (dockyard_rl.experience.cua.environment.OSWorldEnvironment method) global_rope_theta (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) global_rope_type (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) global_segment_size (dockyard_rl.data_plane.interfaces.DataPlaneConfig attribute) global_valid_counts() (in module dockyard_rl.models.jax.train_step) global_valid_seqs (dockyard_rl.algorithms.dpo.DPOValMetrics attribute) global_valid_toks (dockyard_rl.algorithms.dpo.DPOValMetrics attribute) gold_loss (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) GPQADataset (class in dockyard_rl.data.datasets.eval_datasets.gpqa) GPQAEvalDataConfig (class in dockyard_rl.data) gpu_memory_utilization (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) gpu_monitoring (dockyard_rl.utils.logger.LoggerConfig attribute) GpuMetricSnapshot (class in dockyard_rl.utils.logger) GPUMonitoringConfig (class in dockyard_rl.utils.logger) gpus_per_node (dockyard_rl.algorithms.opd.TeacherResourceConfig attribute) (dockyard_rl.cluster.fleet.FleetSpec attribute) (dockyard_rl.distributed.virtual_cluster.ClusterConfig attribute) (dockyard_rl.models.policy.teacher_worker_group.TeacherConfig attribute) gpus_per_server (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) grade() (in module dockyard_rl.rewards.hle_grader) graded_allocation (dockyard_rl.algorithms.grpo.GRPOConfig attribute) graded_allocation_enabled() (in module dockyard_rl.algorithms.rollout_allocation) GRADED_ALLOCATION_GROUP_KEY (in module dockyard_rl.algorithms.rollout_allocation) graded_allocation_metrics() (in module dockyard_rl.algorithms.rollout_allocation) GradedAllocationConfig (class in dockyard_rl.algorithms.rollout_allocation) Grader (in module dockyard_rl.experience.cua.grading.factory) grading_block_network (dockyard_rl.environments.program_bench_environment.ProgramBenchEnvConfig attribute) GradingEnv (class in dockyard_rl.experience.cua.vendor.desktop_env_eval.grading_env) grammar_to_macos() (in module dockyard_rl.experience.cua.actions.cc_transpile) grammar_to_windows() (in module dockyard_rl.experience.cua.actions.cc_transpile) grammar_to_xdotool() (in module dockyard_rl.experience.cua.actions.cc_transpile) ground_truth (dockyard_rl.environments.code_jaccard_environment.CodeJaccardEnvironmentMetadata attribute) (dockyard_rl.environments.math_environment.MathEnvironmentMetadata attribute) (dockyard_rl.environments.retriever.RAGEnvironmentMetadata attribute) (dockyard_rl.environments.vlm_environment.VLMEnvironmentMetadata attribute) group_all_reduce_sum() (in module dockyard_rl.distributed.model_utils) group_all_reduce_sum_with_grad() (in module dockyard_rl.distributed.model_utils) group_and_cat_tensors() (in module dockyard_rl.models.huggingface.common) group_pass_rate() (in module dockyard_rl.algorithms.effort_reward) GroupedExperts (class in dockyard_rl.models.dtensor.moe.experts) (class in dockyard_rl.models.jax.moe.experts) GRPO grpo (dockyard_rl.algorithms.grpo.MasterConfig attribute) grpo_train() (in module dockyard_rl.algorithms.grpo) GRPOAdvantageEstimator (class in dockyard_rl.algorithms.advantage_estimator) GRPOConfig (class in dockyard_rl.algorithms.grpo) GRPOSaveState (class in dockyard_rl.algorithms.grpo) GSM8KDataset (class in dockyard_rl.data.datasets.response_datasets.gsm8k) GSPMD GSPO GuestServerController (class in dockyard_rl.experience.cua.clients.guest_server) (class in dockyard_rl.experience.cua.vendor.desktop_env_eval.grading_env) gym_data_processor() (in module dockyard_rl.data.processors)

H

harness_files (dockyard_rl.sandbox.task_executor_client.TaskSpec attribute) harness_mount (dockyard_rl.sandbox.task_executor_client.TaskSpec attribute) has_valid_call (dockyard_rl.tool_protocol.registry.ToolUseAssessment property) has_violation (dockyard_rl.rewards.invalid_action.InvalidActionVerdict property) head_dim (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) (dockyard_rl.models.jax.models.qwen3.Qwen3Config attribute) (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) (dockyard_rl.models.jax.moe.qwen3_moe.Qwen3MoeConfig attribute) head_dim_for() (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig method) health() (dockyard_rl.experience.cua.vision.TheEyeClient method) helpsteer3_data_processor() (in module dockyard_rl.data.processors) HelpSteer3Dataset (class in dockyard_rl.data.datasets.response_datasets.helpsteer3) HelpSteer3PreferenceDataset (class in dockyard_rl.data.datasets.preference_datasets.helpsteer3) HERMES_TOOL_CALL_END (in module dockyard_rl.tool_protocol.hermes) HERMES_TOOL_CALL_START (in module dockyard_rl.tool_protocol.hermes) HermesExtraction (class in dockyard_rl.tool_protocol.hermes) HF hf_name_to_nnx_path() (in module dockyard_rl.models.jax.weights) HFMultiRewardVerifyWorker (class in dockyard_rl.environments.math_environment)

HFVerifyWorker (class in dockyard_rl.environments.math_environment) hidden_activation (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) hidden_size (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) (dockyard_rl.models.jax.models.qwen3.Qwen3Config attribute) (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) (dockyard_rl.models.jax.moe.qwen3_moe.Qwen3MoeConfig attribute) hidden_size_per_layer_input (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) higher_is_better (dockyard_rl.utils.checkpoint.CheckpointingConfig attribute) HIP_VISIBLE_DEVICES (in module dockyard_rl.utils.device_query) HLE hle_data_processor() (in module dockyard_rl.data.processors) HLEDataset (class in dockyard_rl.data.datasets.hle) HLEEnvConfig (class in dockyard_rl.environments.hle_environment) HLEEnvironment (class in dockyard_rl.environments.hle_environment) HLEJudgeClient (class in dockyard_rl.rewards.hle_grader) host (dockyard_rl.experience.cua.provisioner.HostLease attribute) HostLease (class in dockyard_rl.experience.cua.provisioner) http_server_serving_chat_kwargs (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) HTTPWeightSynchronizer (class in dockyard_rl.weight_sync.http_weight_synchronizer) HV

I

idx (dockyard_rl.data.interfaces.DatumSpec attribute) (dockyard_rl.data.interfaces.KTODatumSpec attribute) (dockyard_rl.data.interfaces.PreferenceDatumSpec attribute) ignore_eos (dockyard_rl.models.generation.interfaces.GenerationConfig attribute) image (dockyard_rl.environments.gdpval_agentic_environment.GDPvalAgenticEnvConfig attribute) (dockyard_rl.sandbox.task_executor_client.SessionStartSpec attribute) (dockyard_rl.sandbox.task_executor_client.TaskResult attribute) (dockyard_rl.sandbox.task_executor_client.TaskSpec attribute) impl (dockyard_rl.data_plane.interfaces.DataPlaneConfig attribute) in_flight_weight_updates (dockyard_rl.algorithms.grpo.AsyncGRPOConfig attribute) INFEASIBLE (dockyard_rl.algorithms.distillation_memory.ColocationSchedule attribute) infeasible() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics) inference_bundle_ct_per_node() (in module dockyard_rl.cluster.placement) inference_sync_ranks() (in module dockyard_rl.cluster.placement) INFINIBAND_SYSFS (in module dockyard_rl.data_plane.rdma) init_collective() (dockyard_rl.models.generation.interfaces.GenerationInterface method) (dockyard_rl.models.generation.sglang.sglang_generation.SGLangGeneration method) (dockyard_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (dockyard_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (dockyard_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) (dockyard_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.interfaces.ColocatablePolicyInterface method) (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) init_communicator() (dockyard_rl.weight_sync.collective_weight_synchronizer.CollectiveWeightSynchronizer method) (dockyard_rl.weight_sync.http_weight_synchronizer.HTTPWeightSynchronizer method) (dockyard_rl.weight_sync.interfaces.WeightSynchronizer method) (dockyard_rl.weight_sync.ipc_weight_synchronizer.IPCWeightSynchronizer method) init_fp8() (in module dockyard_rl.models.generation.vllm.quantization.fp8) init_nccl_communicator() (dockyard_rl.distributed.stateless_process_group.StatelessProcessGroup method) init_ray() (in module dockyard_rl.cluster.bootstrap) init_tmp_checkpoint() (dockyard_rl.utils.checkpoint.CheckpointManager method) input_ids (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossDataDict attribute) (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossDataDict attribute) (dockyard_rl.algorithms.loss.loss_functions.DistillationLossDataDict attribute) (dockyard_rl.algorithms.loss.loss_functions.DPOLossDataDict attribute) (dockyard_rl.algorithms.loss.loss_functions.PreferenceLossDataDict attribute) (dockyard_rl.models.generation.interfaces.GenerationDatumSpec attribute) INPUT_IDS (in module dockyard_rl.data_plane.schema) input_key (dockyard_rl.data.ResponseDatasetConfig attribute) (dockyard_rl.distributed.batched_data_dict.DynamicBatchingArgs attribute) (dockyard_rl.distributed.batched_data_dict.SequencePackingArgs attribute) input_lengths (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossDataDict attribute) (dockyard_rl.algorithms.loss.loss_functions.DistillationLossDataDict attribute) (dockyard_rl.models.generation.interfaces.GenerationDatumSpec attribute) INPUT_LENGTHS (in module dockyard_rl.data_plane.schema) input_lengths_key (dockyard_rl.distributed.batched_data_dict.DynamicBatchingArgs attribute) (dockyard_rl.distributed.batched_data_dict.SequencePackingArgs attribute) input_shape (dockyard_rl.models.dtensor.moe.dispatch.AllToAllDispatchMetadata attribute) input_splits (dockyard_rl.models.dtensor.moe.dispatch.AllToAllDispatchMetadata attribute) input_type (dockyard_rl.algorithms.loss.interfaces.LossFunction attribute) (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossFn attribute) (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn attribute) (dockyard_rl.algorithms.loss.loss_functions.DistillationLossFn attribute) (dockyard_rl.algorithms.loss.loss_functions.DPOLossFn attribute) (dockyard_rl.algorithms.loss.loss_functions.DraftCrossEntropyLossFn attribute) (dockyard_rl.algorithms.loss.loss_functions.NLLLossFn attribute) (dockyard_rl.algorithms.loss.loss_functions.PreferenceLossFn attribute) instruction (dockyard_rl.experience.cua.backends.interface.Observation attribute) integrity_check (dockyard_rl.environments.code_environment.CodeEnvConfig attribute) IntegrityReward (class in dockyard_rl.rewards.integrity) intermediate_size (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) (dockyard_rl.models.jax.models.qwen3.Qwen3Config attribute) (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) (dockyard_rl.models.jax.moe.qwen3_moe.Qwen3MoeConfig attribute) invalid_action (dockyard_rl.rewards.invalid_action.InvalidActionVerdict attribute)

invalid_action_penalty (dockyard_rl.algorithms.grpo.GRPOConfig attribute) (dockyard_rl.rewards.invalid_action.InvalidActionPenaltyConfig attribute) invalid_tool_call_patterns (dockyard_rl.rewards.invalid_action.InvalidActionPenaltyConfig attribute) InvalidActionPenaltyConfig (class in dockyard_rl.rewards.invalid_action) InvalidActionVerdict (class in dockyard_rl.rewards.invalid_action) invalidate_encoder_cache() (dockyard_rl.models.generation.vllm.vllm_generation.VllmGeneration method) invalidate_kv_cache() (dockyard_rl.models.generation.interfaces.GenerationInterface method) (dockyard_rl.models.generation.sglang.sglang_generation.SGLangGeneration method) (dockyard_rl.models.generation.sglang.sglang_worker.SGLangGenerationWorker method) (dockyard_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.interfaces.ColocatablePolicyInterface method) (dockyard_rl.models.policy.lm_policy.Policy method) IPC IPCProtocol (class in dockyard_rl.models.policy.utils) IPCWeightSynchronizer (class in dockyard_rl.weight_sync.ipc_weight_synchronizer) IPOLossFn (class in dockyard_rl.algorithms.loss.loss_functions) IS is_alive() (dockyard_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker method) (dockyard_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) (dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) is_async (dockyard_rl.utils.checkpoint.CheckpointingConfig attribute) is_axis_zero() (dockyard_rl.distributed.named_sharding.NamedSharding static method) is_correct (dockyard_rl.algorithms.x_token.token_aligner.AlignmentPair attribute) is_correct_minerva() (in module dockyard_rl.environments.dapo_math_verifier) is_correct_strict_box() (in module dockyard_rl.environments.dapo_math_verifier) is_fp8_model() (in module dockyard_rl.models.generation.vllm.quantization.fp8) is_fused_moe_module() (in module dockyard_rl.models.generation.vllm.quantization.fp8) is_gemma_model() (in module dockyard_rl.models.huggingface.common) is_gold_text_included_in_pdf() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) is_grouped_expert_key() (in module dockyard_rl.models.dtensor.moe.refit) (in module dockyard_rl.models.jax.moe.refit) is_healthy() (dockyard_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollectorImpl method) is_hf_per_expert_key() (in module dockyard_rl.models.dtensor.moe.convert) is_in_list() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) is_in_vm_clickboard() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.basic_os) is_included_all_json_objects() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) is_kv_shared_layer() (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig method) is_moe_state_dict() (in module dockyard_rl.models.dtensor.moe.detect) is_moe_surgery_supported() (in module dockyard_rl.models.dtensor.moe.surgery) is_non_colocated_teachers_enabled() (in module dockyard_rl.algorithms.opd) is_noop (dockyard_rl.experience.cua.actions.pyautogui_parse.ParsedAction property) is_opd_enabled() (in module dockyard_rl.algorithms.opd) is_peft (dockyard_rl.utils.checkpoint.CheckpointingConfig attribute) is_reference_free() (in module dockyard_rl.algorithms.loss.loss_functions) is_sliding() (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig method) is_sparse_layer() (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig method) (dockyard_rl.models.jax.moe.qwen3_moe.Qwen3MoeConfig method) is_stale (dockyard_rl.weight_sync.collective_weight_synchronizer.CollectiveWeightSynchronizer property) (dockyard_rl.weight_sync.http_weight_synchronizer.HTTPWeightSynchronizer property) (dockyard_rl.weight_sync.interfaces.WeightSynchronizer property) (dockyard_rl.weight_sync.ipc_weight_synchronizer.IPCWeightSynchronizer property) is_terminal (dockyard_rl.experience.cua.actions.pyautogui_parse.ParsedAction property) is_utc_0() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.basic_os) is_vllm_v1_engine_enabled() (in module dockyard_rl.models.policy.utils) item_id (dockyard_rl.rewards.gdpval_rubric.RubricItem attribute) items() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) iter_expanded_refit_tensors() (in module dockyard_rl.models.dtensor.moe.refit) (in module dockyard_rl.models.jax.moe.refit) iter_gemma4_refit_state_dict() (in module dockyard_rl.models.jax.models.gemma4_refit) iter_lb_moe_blocks() (in module dockyard_rl.models.dtensor.moe.load_balance) (in module dockyard_rl.models.jax.moe.load_balance) iter_moe_blocks() (in module dockyard_rl.models.dtensor.moe.router_replay) iter_moe_refit_state_dict() (in module dockyard_rl.models.jax.moe.refit) iter_quant_ignore_name_candidates() (in module dockyard_rl.modelopt.utils) iter_qwen3_next_refit_state_dict() (in module dockyard_rl.models.jax.linear_attn.refit) iter_refit_state_dict() (in module dockyard_rl.models.jax.refit)

J

J0–J11 jax_backend_enabled() (in module dockyard_rl.models.jax.config) JaxPolicyConfig (class in dockyard_rl.models.jax.config) JaxPolicyWorker (class in dockyard_rl.models.jax.policy_worker) JaxPolicyWorkerImpl (class in dockyard_rl.models.jax.policy_worker) jnp_to_torch() (in module dockyard_rl.models.jax.convert) judge() (dockyard_rl.rewards.hle_grader.HLEJudgeClient method) judge_api_key (dockyard_rl.algorithms.online_dpo.OnlineDPOConfig attribute) (dockyard_rl.environments.gdpval_agentic_environment.GDPvalAgenticEnvConfig attribute) (dockyard_rl.environments.gdpval_environment.GDPvalEnvConfig attribute) (dockyard_rl.environments.hle_environment.HLEEnvConfig attribute) judge_base_url (dockyard_rl.algorithms.online_dpo.OnlineDPOConfig attribute) (dockyard_rl.environments.gdpval_agentic_environment.GDPvalAgenticEnvConfig attribute) (dockyard_rl.environments.gdpval_environment.GDPvalEnvConfig attribute) (dockyard_rl.environments.hle_environment.HLEEnvConfig attribute)

judge_model (dockyard_rl.algorithms.online_dpo.OnlineDPOConfig attribute) (dockyard_rl.environments.gdpval_agentic_environment.GDPvalAgenticEnvConfig attribute) (dockyard_rl.environments.gdpval_environment.GDPvalEnvConfig attribute) (dockyard_rl.environments.hle_environment.HLEEnvConfig attribute) judge_pass_rate() (in module dockyard_rl.algorithms.dpo_evals) judge_temperature (dockyard_rl.algorithms.online_dpo.OnlineDPOConfig attribute) judge_timeout (dockyard_rl.algorithms.online_dpo.OnlineDPOConfig attribute) (dockyard_rl.environments.gdpval_agentic_environment.GDPvalAgenticEnvConfig attribute) (dockyard_rl.environments.gdpval_environment.GDPvalEnvConfig attribute) (dockyard_rl.environments.hle_environment.HLEEnvConfig attribute) JudgeOveroptMonitor (class in dockyard_rl.algorithms.online_dpo)

K

KAI kd_loss_mode (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) keep_top_k (dockyard_rl.utils.checkpoint.CheckpointingConfig attribute) keys() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) KL kl_input_clamp_value (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) kl_loss_weight (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) kl_output_clamp_value (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) kl_target (dockyard_rl.algorithms.online_dpo.OnlineDPOConfig attribute) kl_type (dockyard_rl.algorithms.distillation.DistillationConfig attribute) (dockyard_rl.algorithms.loss.loss_functions.DistillationLossConfig attribute)

kto_collate_fn() (in module dockyard_rl.data.collate_fn) KTODatumSpec (class in dockyard_rl.data.interfaces) KTOLossFn (class in dockyard_rl.algorithms.loss.loss_functions) KV cache kv_cache_dtype (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) (dockyard_rl.models.generation.vllm.quantization.fp8.FP8Config attribute) kv_first_write() (in module dockyard_rl.data_plane.column_io) KVBatchMeta (class in dockyard_rl.data_plane.interfaces) KVM

L

l2norm() (in module dockyard_rl.models.jax.linear_attn.delta_rule) label_smoothing (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.loss.loss_functions.DPOLossConfig attribute) last_boxed_only_string() (in module dockyard_rl.environments.dapo_math_verifier) last_put_bytes_per_key (dockyard_rl.data_plane.observability.DataPlaneStats attribute) layer_types (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) layout (dockyard_rl.distributed.named_sharding.NamedSharding property) Layout (in module dockyard_rl.data_plane.schema) length (dockyard_rl.data.interfaces.DatumSpec attribute) (dockyard_rl.data.interfaces.KTODatumSpec attribute) length_chosen (dockyard_rl.data.interfaces.PreferenceDatumSpec attribute) length_penalty (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.loss.loss_functions.DPOLossConfig attribute) length_rejected (dockyard_rl.data.interfaces.PreferenceDatumSpec attribute) Linear (class in dockyard_rl.models.jax.layers) linear_conv_kernel_dim (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) linear_key_head_dim (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) linear_num_key_heads (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) linear_num_value_heads (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) linear_value_head_dim (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) lint_command (dockyard_rl.sandbox.task_executor_client.TaskSpec attribute) lint_errors (dockyard_rl.sandbox.task_executor_client.TaskResult attribute) lint_output (dockyard_rl.sandbox.task_executor_client.TaskResult attribute) literal_match() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) llm (dockyard_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl attribute) LLMMessageLogType (in module dockyard_rl.data.interfaces) load_charts() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.utils) load_checkpoint() (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (in module dockyard_rl.utils.native_checkpoint) load_config() (in module dockyard_rl.utils.config) load_dataset_from_path() (in module dockyard_rl.data.datasets.utils) load_eval_dataset() (in module dockyard_rl.data.datasets.eval_datasets) load_filters() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.utils) load_format (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) load_hf_gemma4_state_dict() (in module dockyard_rl.models.jax.models.gemma4) load_hf_qwen3_moe_state_dict() (in module dockyard_rl.models.jax.moe.weights) load_hf_qwen3_next_state_dict() (in module dockyard_rl.models.jax.linear_attn.weights) load_hf_state_dict() (in module dockyard_rl.models.jax.weights) load_media_from_message() (in module dockyard_rl.data.multimodal_utils) load_pivot_tables() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.utils) load_preference_dataset() (in module dockyard_rl.data.datasets.preference_datasets) load_response_dataset() (in module dockyard_rl.data.datasets.response_datasets) load_rows_or_cols() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.utils) load_sparklines() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.utils) load_state_dict() (dockyard_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) (dockyard_rl.utils.native_checkpoint.ModelState method) (dockyard_rl.utils.native_checkpoint.OptimizerState method) load_training_info() (dockyard_rl.utils.checkpoint.CheckpointManager method) load_weights() (in module dockyard_rl.models.generation.vllm.quantization.fp8) load_xlsx_styles() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.utils) loads_refit_payload() (in module dockyard_rl.models.policy.utils) local (dockyard_rl.models.jax.moe.alltoall.AllToAllDispatchMetadata attribute) local_buffer_size (dockyard_rl.data_plane.interfaces.DataPlaneConfig attribute) local_expert_counts() (in module dockyard_rl.models.jax.moe.alltoall) LocalDispatchMetadata (class in dockyard_rl.models.dtensor.moe.dispatch) (class in dockyard_rl.models.jax.moe.dispatch) localize_alignment() (in module dockyard_rl.algorithms.x_token.loss_utils) LocalizedAlignment (class in dockyard_rl.algorithms.x_token.loss_utils) LocalMathDataset (class in dockyard_rl.data.datasets.eval_datasets.local_math_dataset) LocalMathEvalDataConfig (class in dockyard_rl.data) LocalTokenDispatcher (class in dockyard_rl.models.dtensor.moe.dispatch) (class in dockyard_rl.models.jax.moe.dispatch) locus (dockyard_rl.rewards.invalid_action.Violation attribute) LOCUS_ADVANTAGE (in module dockyard_rl.rewards.invalid_action) locus_overrides (dockyard_rl.rewards.invalid_action.InvalidActionPenaltyConfig attribute) LOCUS_REWARD (in module dockyard_rl.rewards.invalid_action) log_batched_dict_as_jsonl() (dockyard_rl.utils.logger.Logger method) log_dir (dockyard_rl.utils.logger.LoggerConfig attribute) (dockyard_rl.utils.logger.TensorboardConfig attribute) log_event() (in module dockyard_rl.data_plane.observability) log_generation_metrics_to_wandb() (in module dockyard_rl.algorithms.utils) log_histogram() (dockyard_rl.utils.logger.Logger method) (dockyard_rl.utils.logger.LoggerInterface method) (dockyard_rl.utils.logger.MLflowLogger method) (dockyard_rl.utils.logger.SwanlabLogger method) (dockyard_rl.utils.logger.TensorboardLogger method) (dockyard_rl.utils.logger.WandbLogger method) log_hyperparams() (dockyard_rl.utils.logger.Logger method) (dockyard_rl.utils.logger.LoggerInterface method) (dockyard_rl.utils.logger.MLflowLogger method) (dockyard_rl.utils.logger.SwanlabLogger method) (dockyard_rl.utils.logger.TensorboardLogger method) (dockyard_rl.utils.logger.WandbLogger method) log_level (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) log_level_http (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) log_metrics() (dockyard_rl.utils.logger.Logger method) (dockyard_rl.utils.logger.LoggerInterface method) (dockyard_rl.utils.logger.MLflowLogger method) (dockyard_rl.utils.logger.SwanlabLogger method) (dockyard_rl.utils.logger.TensorboardLogger method) (dockyard_rl.utils.logger.WandbLogger method)

log_plot() (dockyard_rl.utils.logger.Logger method) (dockyard_rl.utils.logger.LoggerInterface method) (dockyard_rl.utils.logger.MLflowLogger method) (dockyard_rl.utils.logger.SwanlabLogger method) (dockyard_rl.utils.logger.TensorboardLogger method) (dockyard_rl.utils.logger.WandbLogger method) log_plot_per_worker_timeline_metrics() (dockyard_rl.utils.logger.Logger method) log_plot_token_mult_prob_error() (dockyard_rl.utils.logger.Logger method) log_requests (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) log_requests_level (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) log_string_list_as_jsonl() (dockyard_rl.utils.logger.Logger method) Logger (class in dockyard_rl.utils.logger) logger (dockyard_rl.algorithms.distillation.MasterConfig attribute) (dockyard_rl.algorithms.dpo.MasterConfig attribute) (dockyard_rl.algorithms.grpo.MasterConfig attribute) (dockyard_rl.algorithms.rm.MasterConfig attribute) (dockyard_rl.algorithms.sft.MasterConfig attribute) (in module dockyard_rl.cluster.bootstrap) (in module dockyard_rl.data.multimodal_utils) (in module dockyard_rl.data_plane.adapters.transfer_queue) (in module dockyard_rl.data_plane.observability) (in module dockyard_rl.data_plane.rdma) (in module dockyard_rl.distributed.batched_data_dict) (in module dockyard_rl.distributed.virtual_cluster) (in module dockyard_rl.distributed.worker_groups) (in module dockyard_rl.distributed.worker_groups_utils) (in module dockyard_rl.experience.cua.actions.cc_transpile) (in module dockyard_rl.experience.cua.backends.control_center) (in module dockyard_rl.experience.cua.backends.macos) (in module dockyard_rl.experience.cua.backends.windows) (in module dockyard_rl.experience.cua.clients.control_center) (in module dockyard_rl.experience.cua.clients.guest_server) (in module dockyard_rl.experience.cua.grading.builtins) (in module dockyard_rl.experience.cua.grading.dispatch) (in module dockyard_rl.experience.cua.grading.factory) (in module dockyard_rl.experience.cua.provisioner) (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.dispatch) (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.file) (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.general) (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.info) (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.misc) (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.table) (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.utils) (in module dockyard_rl.experience.rollouts) (in module dockyard_rl.models.generation.sglang.sglang_generation) (in module dockyard_rl.models.generation.sglang.sglang_worker) (in module dockyard_rl.models.generation.vllm.router_capture) (in module dockyard_rl.models.generation.vllm.vllm_generation) (in module dockyard_rl.models.generation.vllm.vllm_worker) (in module dockyard_rl.models.generation.vllm.vllm_worker_async) (in module dockyard_rl.utils.device_capability) (in module dockyard_rl.weight_sync.interfaces) LoggerConfig (class in dockyard_rl.utils.logger) LoggerInterface (class in dockyard_rl.utils.logger) LOGIT (dockyard_rl.algorithms.loss.interfaces.LossInputType attribute) LOGPROB (dockyard_rl.algorithms.loss.interfaces.LossInputType attribute) logprob_batch_size (dockyard_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) LogprobOutputSpec (class in dockyard_rl.models.policy.interfaces) logprobs (dockyard_rl.models.generation.interfaces.GenerationOutputSpec attribute) (dockyard_rl.models.policy.interfaces.LogprobOutputSpec attribute) logprobs_from_logits() (in module dockyard_rl.models.jax.loss) LoRA lora_backend (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) lora_paths (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) lora_target_modules (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) loss (dockyard_rl.algorithms.dpo.DPOValMetrics attribute) (dockyard_rl.algorithms.rm.RMValMetrics attribute) loss_fn (dockyard_rl.algorithms.grpo.MasterConfig attribute) loss_multiplier (dockyard_rl.data.interfaces.DatumSpec attribute) (dockyard_rl.data.interfaces.KTODatumSpec attribute) (dockyard_rl.data.interfaces.PreferenceDatumSpec attribute) loss_type (dockyard_rl.algorithms.loss.interfaces.LossFunction attribute) (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossFn attribute) (dockyard_rl.algorithms.loss.loss_functions.DistillationLossFn attribute) (dockyard_rl.algorithms.loss.loss_functions.DPOLossFn attribute) (dockyard_rl.algorithms.loss.loss_functions.DraftCrossEntropyLossFn attribute) (dockyard_rl.algorithms.loss.loss_functions.NLLLossFn attribute) (dockyard_rl.algorithms.loss.loss_functions.PreferenceLossFn attribute) loss_variant (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.loss.loss_functions.DPOLossConfig attribute) LossFunction (class in dockyard_rl.algorithms.loss.interfaces) LossInputType (class in dockyard_rl.algorithms.loss.interfaces) LossType (class in dockyard_rl.algorithms.loss.interfaces) LP_SEED_FIELDS (in module dockyard_rl.data_plane.schema)

M

MacBackend (class in dockyard_rl.experience.cua.backends.macos) main() (in module dockyard_rl.environments._gdpval_extract) major (dockyard_rl.utils.device_capability.DeviceCapability attribute) make_microbatch_iterator() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) make_microbatch_iterator_for_packable_sequences() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) make_microbatch_iterator_with_dynamic_shapes() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) make_token_dispatcher() (in module dockyard_rl.models.jax.moe.alltoall) make_tool() (in module dockyard_rl.tool_protocol.registry) malformed (dockyard_rl.tool_protocol.registry.ToolUseAssessment attribute) malformed_thinking (dockyard_rl.rewards.invalid_action.InvalidActionVerdict attribute) malformed_thinking_penalty (dockyard_rl.rewards.invalid_action.InvalidActionPenaltyConfig attribute) manual() (in module dockyard_rl.experience.cua.grading.builtins) map() (dockyard_rl.data.datasets.response_datasets.oai_format_dataset.PreservingDataset method) mark_iteration() (dockyard_rl.utils.timer.TimeoutChecker method) mark_stale() (dockyard_rl.weight_sync.collective_weight_synchronizer.CollectiveWeightSynchronizer method) (dockyard_rl.weight_sync.http_weight_synchronizer.HTTPWeightSynchronizer method) (dockyard_rl.weight_sync.interfaces.WeightSynchronizer method) (dockyard_rl.weight_sync.ipc_weight_synchronizer.IPCWeightSynchronizer method) mask_out_neg_inf_logprobs() (in module dockyard_rl.algorithms.utils) masked_mean() (in module dockyard_rl.algorithms.loss.utils) (in module dockyard_rl.models.jax.loss) master_port_range_high (dockyard_rl.distributed.virtual_cluster.ClusterConfig attribute) master_port_range_low (dockyard_rl.distributed.virtual_cluster.ClusterConfig attribute) MasterConfig (class in dockyard_rl.algorithms.distillation) (class in dockyard_rl.algorithms.dpo) (class in dockyard_rl.algorithms.grpo) (class in dockyard_rl.algorithms.rm) (class in dockyard_rl.algorithms.sft) match_in_list() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) matches() (dockyard_rl.models.huggingface.common.ModelFlag method) matches_quant_ignore_pattern() (in module dockyard_rl.modelopt.utils) materialize() (in module dockyard_rl.data_plane.codec) math_data_processor() (in module dockyard_rl.data.processors) math_expression_reward() (in module dockyard_rl.environments.rewards) math_hf_data_processor() (in module dockyard_rl.data.processors) math_verify_func (in module dockyard_rl.environments.rewards) math_verify_impl (dockyard_rl.environments.math_environment.MathEnvConfig attribute) MathDataset (class in dockyard_rl.data.datasets.eval_datasets.math) MathEnvConfig (class in dockyard_rl.environments.math_environment) MathEnvironment (class in dockyard_rl.environments.math_environment) MathEnvironmentMetadata (class in dockyard_rl.environments.math_environment) MathEvalDataConfig (class in dockyard_rl.data) MathMultiRewardEnvironment (class in dockyard_rl.environments.math_environment) max_bytes_per_key_seen (dockyard_rl.data_plane.observability.DataPlaneStats attribute) max_colocated_worker_groups (dockyard_rl.cluster.fleet.FleetSpec attribute) max_concurrency (dockyard_rl.environments.gdpval_agentic_environment.GDPvalAgenticEnvConfig attribute) (dockyard_rl.environments.gdpval_environment.GDPvalEnvConfig attribute) (dockyard_rl.environments.hle_environment.HLEEnvConfig attribute) (dockyard_rl.environments.program_bench_environment.ProgramBenchEnvConfig attribute) (dockyard_rl.environments.terminal_bench_environment.TerminalBenchEnvConfig attribute) max_doc_length (dockyard_rl.environments.retriever.RAGEnvConfig attribute) max_generation_failures (dockyard_rl.algorithms.grpo.AsyncGRPOConfig attribute) max_grad_norm (dockyard_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) max_image_pixels (dockyard_rl.models.generation.interfaces.GenerationConfig attribute) max_images_per_sample (dockyard_rl.models.generation.interfaces.GenerationConfig attribute) max_input_seq_length (dockyard_rl.data.AIMEEvalDataConfig attribute) (dockyard_rl.data.DataConfig attribute) (dockyard_rl.data.GPQAEvalDataConfig attribute) (dockyard_rl.data.LocalMathEvalDataConfig attribute) (dockyard_rl.data.MathEvalDataConfig attribute) (dockyard_rl.data.MMAUEvalDataConfig attribute) (dockyard_rl.data.MMMLUEvalDataConfig attribute) (dockyard_rl.data.MMMLUProEvalDataConfig attribute) max_loaded_loras (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) max_lora_rank (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) max_loras_per_batch (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) max_model_len (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) max_new_tokens (dockyard_rl.models.generation.interfaces.GenerationConfig attribute) max_num_epochs (dockyard_rl.algorithms.distillation.DistillationConfig attribute) (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.grpo.GRPOConfig attribute) (dockyard_rl.algorithms.rm.RMConfig attribute) (dockyard_rl.algorithms.sft.SFTConfig attribute) max_num_steps (dockyard_rl.algorithms.distillation.DistillationConfig attribute) (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.grpo.GRPOConfig attribute) (dockyard_rl.algorithms.online_dpo.OnlineDPOConfig attribute) (dockyard_rl.algorithms.rm.RMConfig attribute) (dockyard_rl.algorithms.sft.SFTConfig attribute) max_pack_size (dockyard_rl.data.packing.metrics.PackingMetrics attribute) (dockyard_rl.data.packing.packing.PackingMetrics attribute) max_prefill_tokens (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) max_response_length (dockyard_rl.algorithms.reward_functions.RewardShapingConfig attribute) max_response_tokens (dockyard_rl.algorithms.effort_reward.EffortRewardConfig attribute) max_rollout_turns (dockyard_rl.algorithms.grpo.GRPOConfig attribute) max_running_requests (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) max_seqlen_k (dockyard_rl.models.huggingface.common.FlashAttentionKwargs attribute) max_seqlen_q (dockyard_rl.models.huggingface.common.FlashAttentionKwargs attribute) max_tokens_per_microbatch (dockyard_rl.distributed.batched_data_dict.DynamicBatchingArgs attribute) (dockyard_rl.distributed.batched_data_dict.SequencePackingArgs attribute) max_trajectory_age_steps (dockyard_rl.algorithms.grpo.AsyncGRPOConfig attribute) max_turns (dockyard_rl.environments.gdpval_agentic_environment.GDPvalAgenticEnvConfig attribute) (dockyard_rl.environments.program_bench_environment.ProgramBenchEnvConfig attribute) (dockyard_rl.environments.terminal_bench_environment.TerminalBenchEnvConfig attribute) max_val_samples (dockyard_rl.algorithms.grpo.GRPOConfig attribute) maybe_gpu_profile_step() (in module dockyard_rl.utils.nsys) maybe_init_zmq() (dockyard_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) maybe_pack_jagged() (in module dockyard_rl.data_plane.codec) maybe_pad_last_batch() (in module dockyard_rl.algorithms.utils) maybe_post_process_fp8_weight_block() (in module dockyard_rl.models.generation.vllm.quantization.fp8) mean_valid_logprob() (in module dockyard_rl.algorithms.teacher_panel) MEDIA_TAG_PATTERN (in module dockyard_rl.data.multimodal_utils) MEDIA_TAGS (in module dockyard_rl.data.multimodal_utils) MEDIA_TAGS_REVERSED (in module dockyard_rl.data.multimodal_utils) MEDIA_TAGS_TO_ALLOWED (in module dockyard_rl.data.multimodal_utils) mem_fraction_static (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) mem_used_diff_gb (dockyard_rl.utils.memory_tracker.MemoryTrackerDataPoint property) memory_info() (dockyard_rl.utils.device_query.DeviceQueryBackend method) (dockyard_rl.utils.device_query.NvmlBackend method) memory_used_after_stage_gb (dockyard_rl.utils.memory_tracker.MemoryTrackerDataPoint attribute) memory_used_before_stage_gb (dockyard_rl.utils.memory_tracker.MemoryTrackerDataPoint attribute) MemoryReport (class in dockyard_rl.algorithms.distillation_memory) MemoryTracker (class in dockyard_rl.utils.memory_tracker) MemoryTrackerDataPoint (class in dockyard_rl.utils.memory_tracker) merge_fp8_kwargs() (in module dockyard_rl.models.generation.vllm.vllm_worker) MESH_AXES (in module dockyard_rl.models.jax.sharding) MESH_AXES_MOE (in module dockyard_rl.models.jax.sharding) message_log (dockyard_rl.data.interfaces.DatumSpec attribute) (dockyard_rl.data.interfaces.KTODatumSpec attribute) MESSAGE_LOG_BULK_FIELDS (in module dockyard_rl.data.llm_message_utils) message_log_chosen (dockyard_rl.data.interfaces.PreferenceDatumSpec attribute) message_log_rejected (dockyard_rl.data.interfaces.PreferenceDatumSpec attribute) message_log_shape() (in module dockyard_rl.data.llm_message_utils) message_log_to_flat_messages() (in module dockyard_rl.data.llm_message_utils) META_IDX (in module dockyard_rl.data_plane.schema) metadata (dockyard_rl.environments.interfaces.EnvironmentReturn attribute) MetadataT (in module dockyard_rl.environments.interfaces) metric_name (dockyard_rl.utils.checkpoint.CheckpointingConfig attribute) metrics (dockyard_rl.utils.logger.GpuMetricSnapshot attribute) MetricsDataPlaneClient (class in dockyard_rl.data_plane.observability) MICRO_BATCH_INDICES (in module dockyard_rl.data_plane.schema) MICRO_BATCH_LENGTHS (in module dockyard_rl.data_plane.schema) micro_batch_size (dockyard_rl.algorithms.opd.TeacherResourceConfig attribute) (dockyard_rl.models.policy.teacher_worker_group.TeacherConfig attribute) min_rollouts_per_prompt (dockyard_rl.algorithms.rollout_allocation.GradedAllocationConfig attribute) minor (dockyard_rl.utils.device_capability.DeviceCapability attribute) minus_baseline (dockyard_rl.algorithms.grpo.AdvEstimatorConfig attribute) MISSING_ROUTE_SENTINEL (in module dockyard_rl.models.dtensor.moe.router) (in module dockyard_rl.models.generation.vllm.router_capture) mixed_kl_weight (dockyard_rl.algorithms.distillation.DistillationConfig attribute) (dockyard_rl.algorithms.loss.loss_functions.DistillationLossConfig attribute) mlflow (dockyard_rl.utils.logger.LoggerConfig attribute) mlflow_enabled (dockyard_rl.utils.logger.LoggerConfig attribute) MLflowConfig (class in dockyard_rl.utils.logger) MLflowLogger (class in dockyard_rl.utils.logger) mlp_intermediate_size() (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig method) mlp_only_layers (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) (dockyard_rl.models.jax.moe.qwen3_moe.Qwen3MoeConfig attribute) MMAUDataset (class in dockyard_rl.data.datasets.eval_datasets.mmau) MMAUEvalDataConfig (class in dockyard_rl.data) MMLUDataset (class in dockyard_rl.data.datasets.eval_datasets.mmlu) MMLUProDataset (class in dockyard_rl.data.datasets.eval_datasets.mmlu_pro) MMMLUEvalDataConfig (class in dockyard_rl.data) MMMLUProEvalDataConfig (class in dockyard_rl.data) mode (dockyard_rl.sandbox.task_executor_client.SessionStartSpec attribute) (dockyard_rl.sandbox.task_executor_client.TaskResult attribute) (dockyard_rl.sandbox.task_executor_client.TaskSpec attribute) model_cache_dir (dockyard_rl.utils.checkpoint.CheckpointingConfig attribute) model_name (dockyard_rl.algorithms.distillation.TeacherConfig attribute) (dockyard_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) (dockyard_rl.models.generation.interfaces.GenerationConfig attribute) (dockyard_rl.models.policy.teacher_worker_group.TeacherConfig attribute) model_parallel_size (dockyard_rl.cluster.placement.ParallelismSpec property) (dockyard_rl.models.generation.vllm.quantization.fp8.FP8Config attribute) model_path (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) model_post_init() (dockyard_rl.utils.memory_tracker.MemoryTracker method) model_repo_id (dockyard_rl.utils.checkpoint.CheckpointingConfig attribute) model_save_format (dockyard_rl.utils.checkpoint.CheckpointingConfig attribute) ModelFlag (class in dockyard_rl.models.huggingface.common) MODELOPT_ACTOR_REGISTRY (in module dockyard_rl.modelopt.registry) ModelState (class in dockyard_rl.utils.native_checkpoint) module dockyard_rl dockyard_rl.algorithms dockyard_rl.algorithms.advantage_estimator dockyard_rl.algorithms.async_utils dockyard_rl.algorithms.async_utils.interfaces dockyard_rl.algorithms.async_utils.replay_buffer dockyard_rl.algorithms.async_utils.trajectory_collector dockyard_rl.algorithms.config_access dockyard_rl.algorithms.distillation dockyard_rl.algorithms.distillation_memory dockyard_rl.algorithms.dpo dockyard_rl.algorithms.dpo_evals dockyard_rl.algorithms.effort_reward dockyard_rl.algorithms.grpo dockyard_rl.algorithms.logits_sampling_utils dockyard_rl.algorithms.loss dockyard_rl.algorithms.loss.interfaces dockyard_rl.algorithms.loss.loss_functions dockyard_rl.algorithms.loss.utils dockyard_rl.algorithms.loss.wrapper dockyard_rl.algorithms.online_dpo dockyard_rl.algorithms.opd dockyard_rl.algorithms.reward_functions dockyard_rl.algorithms.rlaif dockyard_rl.algorithms.rm dockyard_rl.algorithms.rollout_allocation dockyard_rl.algorithms.routing_diversity dockyard_rl.algorithms.sft dockyard_rl.algorithms.teacher_panel dockyard_rl.algorithms.utils dockyard_rl.algorithms.x_token dockyard_rl.algorithms.x_token.loss_utils dockyard_rl.algorithms.x_token.token_aligner dockyard_rl.algorithms.x_token.utils dockyard_rl.cluster dockyard_rl.cluster.bootstrap dockyard_rl.cluster.fleet dockyard_rl.cluster.placement dockyard_rl.data dockyard_rl.data.chat_templates dockyard_rl.data.collate_fn dockyard_rl.data.cross_tokenizer_collate dockyard_rl.data.dataloader dockyard_rl.data.datasets dockyard_rl.data.datasets.eval_datasets dockyard_rl.data.datasets.eval_datasets.aime dockyard_rl.data.datasets.eval_datasets.gpqa dockyard_rl.data.datasets.eval_datasets.local_math_dataset dockyard_rl.data.datasets.eval_datasets.math dockyard_rl.data.datasets.eval_datasets.mmau dockyard_rl.data.datasets.eval_datasets.mmlu dockyard_rl.data.datasets.eval_datasets.mmlu_pro dockyard_rl.data.datasets.gdpval dockyard_rl.data.datasets.hle dockyard_rl.data.datasets.preference_datasets dockyard_rl.data.datasets.preference_datasets.binary_preference_dataset dockyard_rl.data.datasets.preference_datasets.helpsteer3 dockyard_rl.data.datasets.preference_datasets.preference_dataset dockyard_rl.data.datasets.preference_datasets.tulu3 dockyard_rl.data.datasets.processed_dataset dockyard_rl.data.datasets.program_bench dockyard_rl.data.datasets.raw_dataset dockyard_rl.data.datasets.response_datasets dockyard_rl.data.datasets.response_datasets.aime24 dockyard_rl.data.datasets.response_datasets.avqa dockyard_rl.data.datasets.response_datasets.clevr dockyard_rl.data.datasets.response_datasets.daily_omni dockyard_rl.data.datasets.response_datasets.dapo_math dockyard_rl.data.datasets.response_datasets.deepscaler dockyard_rl.data.datasets.response_datasets.gemometry3k dockyard_rl.data.datasets.response_datasets.general_conversation_dataset dockyard_rl.data.datasets.response_datasets.gsm8k dockyard_rl.data.datasets.response_datasets.helpsteer3 dockyard_rl.data.datasets.response_datasets.oai_format_dataset dockyard_rl.data.datasets.response_datasets.oasst dockyard_rl.data.datasets.response_datasets.openmathinstruct2 dockyard_rl.data.datasets.response_datasets.refcoco dockyard_rl.data.datasets.response_datasets.response_dataset dockyard_rl.data.datasets.response_datasets.squad dockyard_rl.data.datasets.response_datasets.tulu3 dockyard_rl.data.datasets.swe_bench dockyard_rl.data.datasets.swe_bench_pro dockyard_rl.data.datasets.terminal_bench dockyard_rl.data.datasets.utils dockyard_rl.data.interfaces dockyard_rl.data.llm_message_utils dockyard_rl.data.multimodal_utils dockyard_rl.data.packing dockyard_rl.data.packing.algorithms dockyard_rl.data.packing.metrics dockyard_rl.data.packing.packing dockyard_rl.data.processors dockyard_rl.data.utils dockyard_rl.data_plane dockyard_rl.data_plane.adapters dockyard_rl.data_plane.adapters.noop dockyard_rl.data_plane.adapters.transfer_queue dockyard_rl.data_plane.codec dockyard_rl.data_plane.column_io dockyard_rl.data_plane.factory dockyard_rl.data_plane.interfaces dockyard_rl.data_plane.observability dockyard_rl.data_plane.preshard dockyard_rl.data_plane.rdma dockyard_rl.data_plane.schema dockyard_rl.data_plane.worker_mixin dockyard_rl.distributed dockyard_rl.distributed.batched_data_dict dockyard_rl.distributed.model_utils dockyard_rl.distributed.named_sharding dockyard_rl.distributed.stateless_process_group dockyard_rl.distributed.virtual_cluster dockyard_rl.distributed.worker_groups dockyard_rl.distributed.worker_groups_utils dockyard_rl.environments dockyard_rl.environments._gdpval_extract dockyard_rl.environments.code_environment dockyard_rl.environments.code_jaccard_environment dockyard_rl.environments.dapo_math_verifier dockyard_rl.environments.gdpval_agentic_environment dockyard_rl.environments.gdpval_environment dockyard_rl.environments.hle_environment dockyard_rl.environments.interfaces dockyard_rl.environments.math_environment dockyard_rl.environments.metrics dockyard_rl.environments.multi_turn_session_env dockyard_rl.environments.program_bench_environment dockyard_rl.environments.retriever dockyard_rl.environments.reward_model_environment dockyard_rl.environments.rewards dockyard_rl.environments.terminal_bench_environment dockyard_rl.environments.utils dockyard_rl.environments.vlm_environment dockyard_rl.evals dockyard_rl.evals.answer_parsing dockyard_rl.experience dockyard_rl.experience.cua dockyard_rl.experience.cua.actions dockyard_rl.experience.cua.actions.cc_transpile dockyard_rl.experience.cua.actions.pyautogui_parse dockyard_rl.experience.cua.backends dockyard_rl.experience.cua.backends.control_center dockyard_rl.experience.cua.backends.interface dockyard_rl.experience.cua.backends.macos dockyard_rl.experience.cua.backends.official dockyard_rl.experience.cua.backends.windows dockyard_rl.experience.cua.clients dockyard_rl.experience.cua.clients.control_center dockyard_rl.experience.cua.clients.guest_server dockyard_rl.experience.cua.clients.proto dockyard_rl.experience.cua.datasets dockyard_rl.experience.cua.datasets.osworld dockyard_rl.experience.cua.environment dockyard_rl.experience.cua.grading dockyard_rl.experience.cua.grading.builtins dockyard_rl.experience.cua.grading.dispatch dockyard_rl.experience.cua.grading.factory dockyard_rl.experience.cua.grading.registry dockyard_rl.experience.cua.provisioner dockyard_rl.experience.cua.rollout dockyard_rl.experience.cua.tasks dockyard_rl.experience.cua.tasks.validate dockyard_rl.experience.cua.vendor dockyard_rl.experience.cua.vendor.desktop_env_eval dockyard_rl.experience.cua.vendor.desktop_env_eval.dispatch dockyard_rl.experience.cua.vendor.desktop_env_eval.getters dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.file dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.general dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.info dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.misc dockyard_rl.experience.cua.vendor.desktop_env_eval.grading_env dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.basic_os dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.pdf dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.table dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.utils dockyard_rl.experience.cua.vision dockyard_rl.experience.rollouts dockyard_rl.modelopt dockyard_rl.modelopt.registry dockyard_rl.modelopt.utils dockyard_rl.models dockyard_rl.models.dtensor dockyard_rl.models.dtensor.moe dockyard_rl.models.dtensor.moe.block dockyard_rl.models.dtensor.moe.config dockyard_rl.models.dtensor.moe.convert dockyard_rl.models.dtensor.moe.detect dockyard_rl.models.dtensor.moe.dispatch dockyard_rl.models.dtensor.moe.experts dockyard_rl.models.dtensor.moe.load_balance dockyard_rl.models.dtensor.moe.mesh dockyard_rl.models.dtensor.moe.parallelize_moe dockyard_rl.models.dtensor.moe.refit dockyard_rl.models.dtensor.moe.router dockyard_rl.models.dtensor.moe.router_replay dockyard_rl.models.dtensor.moe.sharding dockyard_rl.models.dtensor.moe.surgery dockyard_rl.models.dtensor.parallelize dockyard_rl.models.generation dockyard_rl.models.generation.constants dockyard_rl.models.generation.interfaces dockyard_rl.models.generation.sglang dockyard_rl.models.generation.sglang.config dockyard_rl.models.generation.sglang.sglang_copied_utils dockyard_rl.models.generation.sglang.sglang_generation dockyard_rl.models.generation.sglang.sglang_worker dockyard_rl.models.generation.sglang.utils dockyard_rl.models.generation.vllm dockyard_rl.models.generation.vllm.config dockyard_rl.models.generation.vllm.quantization dockyard_rl.models.generation.vllm.quantization.fp8 dockyard_rl.models.generation.vllm.quantization.fp8_train_utils dockyard_rl.models.generation.vllm.router_capture dockyard_rl.models.generation.vllm.utils dockyard_rl.models.generation.vllm.vllm_backend dockyard_rl.models.generation.vllm.vllm_generation dockyard_rl.models.generation.vllm.vllm_worker dockyard_rl.models.generation.vllm.vllm_worker_async dockyard_rl.models.huggingface dockyard_rl.models.huggingface.common dockyard_rl.models.jax dockyard_rl.models.jax.checkpoint dockyard_rl.models.jax.config dockyard_rl.models.jax.convert dockyard_rl.models.jax.layers dockyard_rl.models.jax.linear_attn dockyard_rl.models.jax.linear_attn.delta_rule dockyard_rl.models.jax.linear_attn.gated_deltanet dockyard_rl.models.jax.linear_attn.refit dockyard_rl.models.jax.linear_attn.weights dockyard_rl.models.jax.loss dockyard_rl.models.jax.models dockyard_rl.models.jax.models.gemma4 dockyard_rl.models.jax.models.gemma4_refit dockyard_rl.models.jax.models.qwen3 dockyard_rl.models.jax.models.qwen3_next dockyard_rl.models.jax.moe dockyard_rl.models.jax.moe.alltoall dockyard_rl.models.jax.moe.block dockyard_rl.models.jax.moe.dispatch dockyard_rl.models.jax.moe.experts dockyard_rl.models.jax.moe.load_balance dockyard_rl.models.jax.moe.qwen3_moe dockyard_rl.models.jax.moe.refit dockyard_rl.models.jax.moe.router dockyard_rl.models.jax.moe.weights dockyard_rl.models.jax.optimizer dockyard_rl.models.jax.policy_worker dockyard_rl.models.jax.refit dockyard_rl.models.jax.sharding dockyard_rl.models.jax.train_step dockyard_rl.models.jax.weights dockyard_rl.models.policy dockyard_rl.models.policy.interfaces dockyard_rl.models.policy.lm_policy dockyard_rl.models.policy.teacher_scorer dockyard_rl.models.policy.teacher_worker_group dockyard_rl.models.policy.tq_policy dockyard_rl.models.policy.utils dockyard_rl.models.policy.workers dockyard_rl.models.policy.workers.base_policy_worker dockyard_rl.models.policy.workers.dtensor_policy_worker dockyard_rl.models.policy.workers.dtensor_policy_worker_v2 dockyard_rl.rewards dockyard_rl.rewards.compiler dockyard_rl.rewards.diff_utils dockyard_rl.rewards.gdpval_rubric dockyard_rl.rewards.hle_grader dockyard_rl.rewards.integrity dockyard_rl.rewards.interfaces dockyard_rl.rewards.invalid_action dockyard_rl.rewards.program_bench dockyard_rl.rewards.swe_bench_pro dockyard_rl.rewards.terminal_bench dockyard_rl.rewards.test_runner dockyard_rl.sandbox dockyard_rl.sandbox.task_executor_client dockyard_rl.tool_protocol dockyard_rl.tool_protocol.hermes dockyard_rl.tool_protocol.protocol dockyard_rl.tool_protocol.registry dockyard_rl.tool_protocol.schema dockyard_rl.utils dockyard_rl.utils.checkpoint dockyard_rl.utils.config dockyard_rl.utils.device_capability dockyard_rl.utils.device_query dockyard_rl.utils.gguf_export dockyard_rl.utils.logger dockyard_rl.utils.memory_tracker dockyard_rl.utils.native_checkpoint dockyard_rl.utils.nsys dockyard_rl.utils.packed_tensor dockyard_rl.utils.timer dockyard_rl.weight_sync dockyard_rl.weight_sync.collective_weight_synchronizer dockyard_rl.weight_sync.factory dockyard_rl.weight_sync.http_weight_synchronizer dockyard_rl.weight_sync.interfaces dockyard_rl.weight_sync.ipc_weight_synchronizer

MoE moe_dims (dockyard_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl attribute) moe_intermediate_size (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) (dockyard_rl.models.jax.moe.qwen3_moe.Qwen3MoeConfig attribute) moe_load_balance_coeff (dockyard_rl.models.jax.config.JaxPolicyConfig attribute) moe_mesh (dockyard_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl attribute) moe_token_dispatcher (dockyard_rl.models.jax.config.JaxPolicyConfig attribute) MoEBlock (class in dockyard_rl.models.dtensor.moe.block) (class in dockyard_rl.models.jax.moe.block) MoEMeshDims (class in dockyard_rl.models.jax.sharding) MoEParallelDims (class in dockyard_rl.models.dtensor.moe.mesh) monitor_gpus (dockyard_rl.utils.logger.LoggerConfig attribute) monkey_patch_torch_reductions() (in module dockyard_rl.models.generation.sglang.sglang_copied_utils) monkey_patch_vllm_ray_executor() (in module dockyard_rl.models.generation.vllm.quantization.fp8) month_mapping (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.misc) Month_Mapping_Full (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.misc) month_mapping_full (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.misc) move_buffer_to_device() (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) move_optimizer_to_device() (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) move_to_cpu() (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) move_to_cuda() (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) move_to_device() (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) multichoice_qa_processor() (in module dockyard_rl.data.processors) MULTILINGUAL_ANSWER_PATTERN_TEMPLATE (in module dockyard_rl.evals.answer_parsing) MULTILINGUAL_ANSWER_REGEXES (in module dockyard_rl.evals.answer_parsing) MultilingualMultichoiceVerifyWorker (class in dockyard_rl.environments.math_environment) MultipleDataloaderWrapper (class in dockyard_rl.data.dataloader) MultiprocessingSerializer (class in dockyard_rl.models.generation.sglang.sglang_copied_utils) MultiTurnSessionEnvironment (class in dockyard_rl.environments.multi_turn_session_env) MultiWorkerFuture (class in dockyard_rl.distributed.worker_groups) my_init() (in module dockyard_rl.models.generation.vllm.quantization.fp8) my_run_engine_core() (in module dockyard_rl.models.generation.vllm.quantization.fp8)

N

n_bytes (dockyard_rl.data_plane.observability.DataPlaneEvent attribute) n_keys (dockyard_rl.data_plane.observability.DataPlaneEvent attribute) name (dockyard_rl.algorithms.grpo.AdvEstimatorConfig attribute) (dockyard_rl.algorithms.rlaif.ConstitutionPrinciple attribute) (dockyard_rl.cluster.fleet.FleetSpec attribute) (dockyard_rl.tool_protocol.hermes.ParsedToolCall attribute) (dockyard_rl.utils.device_capability.DeviceCapability attribute) (dockyard_rl.utils.device_query.DeviceQueryBackend attribute) (dockyard_rl.utils.device_query.NvmlBackend attribute) (dockyard_rl.utils.logger.SwanlabConfig attribute) (dockyard_rl.utils.logger.WandbConfig attribute) NamedSharding (class in dockyard_rl.distributed.named_sharding) names (dockyard_rl.distributed.named_sharding.NamedSharding property) names() (dockyard_rl.tool_protocol.registry.ToolRegistry method) NCCL nccl_sync_world_size() (in module dockyard_rl.cluster.placement) ndim (dockyard_rl.distributed.named_sharding.NamedSharding property) need_top_k_or_top_p_filtering() (in module dockyard_rl.algorithms.logits_sampling_utils) new_variables (dockyard_rl.utils.memory_tracker.MemoryTrackerDataPoint property) next_stop_strings (dockyard_rl.environments.interfaces.EnvironmentReturn attribute) next_token_accuracy() (in module dockyard_rl.algorithms.x_token.loss_utils) NLL NLLLossFn (class in dockyard_rl.algorithms.loss.loss_functions) NNX nnx_path_to_hf_name() (in module dockyard_rl.models.jax.weights) node_count() (dockyard_rl.distributed.virtual_cluster.RayVirtualCluster method) non_colocated_teachers (dockyard_rl.algorithms.opd.OnPolicyDistillationConfig attribute) NonColocatedTeachersConfig (class in dockyard_rl.algorithms.opd) NoOpDataPlaneClient (class in dockyard_rl.data_plane.adapters.noop) norm_topk_prob (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) (dockyard_rl.models.jax.moe.qwen3_moe.Qwen3MoeConfig attribute) normalize() (in module dockyard_rl.rewards.hle_grader) normalize_and_validate_image() (in module dockyard_rl.data.multimodal_utils) normalize_effort_cost() (in module dockyard_rl.algorithms.effort_reward) normalize_extracted_answer() (in module dockyard_rl.evals.answer_parsing) normalize_final_answer() (in module dockyard_rl.environments.dapo_math_verifier) normalize_response() (in module dockyard_rl.evals.answer_parsing) normalize_rewards (dockyard_rl.algorithms.grpo.AdvEstimatorConfig attribute) (dockyard_rl.algorithms.grpo.GRPOConfig attribute) normalize_teacher_by_vocab (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) NOSET_VISIBLE_DEVICE_ENV_VARS (in module dockyard_rl.utils.device_query) num_attention_heads (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) (dockyard_rl.models.jax.models.qwen3.Qwen3Config attribute) (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) (dockyard_rl.models.jax.moe.qwen3_moe.Qwen3MoeConfig attribute) num_chunks (dockyard_rl.algorithms.x_token.token_aligner.AlignmentBatch attribute)

num_continuous_decode_steps (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) num_experts (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) (dockyard_rl.models.jax.moe.qwen3_moe.Qwen3MoeConfig attribute) num_experts_per_tok (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) (dockyard_rl.models.jax.moe.qwen3_moe.Qwen3MoeConfig attribute) num_first_layers_in_bf16 (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) (dockyard_rl.models.generation.vllm.quantization.fp8.FP8Config attribute) num_generations_per_prompt (dockyard_rl.algorithms.grpo.GRPOConfig attribute) num_global_key_value_heads (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) num_hidden_layers (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) (dockyard_rl.models.jax.models.qwen3.Qwen3Config attribute) (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) (dockyard_rl.models.jax.moe.qwen3_moe.Qwen3MoeConfig attribute) num_key_value_heads (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) (dockyard_rl.models.jax.models.qwen3.Qwen3Config attribute) (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) (dockyard_rl.models.jax.moe.qwen3_moe.Qwen3MoeConfig attribute) num_kv_shared_layers (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) num_last_layers_in_bf16 (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) (dockyard_rl.models.generation.vllm.quantization.fp8.FP8Config attribute) num_local_experts() (dockyard_rl.models.dtensor.moe.mesh.MoEParallelDims method) (dockyard_rl.models.jax.sharding.MoEMeshDims method) (in module dockyard_rl.models.dtensor.moe.sharding) num_nodes (dockyard_rl.algorithms.opd.TeacherResourceConfig attribute) (dockyard_rl.cluster.fleet.FleetSpec attribute) (dockyard_rl.distributed.virtual_cluster.ClusterConfig attribute) (dockyard_rl.models.policy.teacher_worker_group.TeacherConfig attribute) num_prompts_per_dataloader (dockyard_rl.data.DataConfig attribute) num_prompts_per_step (dockyard_rl.algorithms.grpo.GRPOConfig attribute) num_storage_units (dockyard_rl.data_plane.interfaces.DataPlaneConfig attribute) num_threads (dockyard_rl.environments.gdpval_environment.GDPvalEnvConfig attribute) (dockyard_rl.environments.hle_environment.HLEEnvConfig attribute) num_topk_logits (dockyard_rl.algorithms.distillation.DistillationConfig attribute) num_val_samples_to_print (dockyard_rl.utils.logger.LoggerConfig attribute) num_valid_samples (dockyard_rl.algorithms.dpo.DPOValMetrics attribute) (dockyard_rl.algorithms.rm.RMValMetrics attribute) num_workers (dockyard_rl.data.DataConfig attribute) (dockyard_rl.environments.code_environment.CodeEnvConfig attribute) (dockyard_rl.environments.code_jaccard_environment.CodeJaccardEnvConfig attribute) (dockyard_rl.environments.math_environment.MathEnvConfig attribute) (dockyard_rl.environments.retriever.RAGEnvConfig attribute) (dockyard_rl.environments.vlm_environment.VLMEnvConfig attribute) NVFP4 NVLINK_DOMAIN_PREFIX (in module dockyard_rl.distributed.virtual_cluster) NVLINK_DOMAIN_UNKNOWN (in module dockyard_rl.distributed.virtual_cluster) NvmlBackend (class in dockyard_rl.utils.device_query)

O

OasstDataset (class in dockyard_rl.data.datasets.response_datasets.oasst) observability (dockyard_rl.data_plane.interfaces.DataPlaneConfig attribute) ObservabilityConfig (class in dockyard_rl.data_plane.interfaces) Observation (class in dockyard_rl.experience.cua.backends.interface) observations (dockyard_rl.environments.interfaces.EnvironmentReturn attribute) observe() (dockyard_rl.experience.cua.backends.control_center.ControlCenterBackend method) (dockyard_rl.experience.cua.backends.interface.OSWorldBackend method) (dockyard_rl.experience.cua.backends.macos.MacBackend method) (dockyard_rl.experience.cua.backends.official.OfficialDesktopEnvBackend method) (dockyard_rl.experience.cua.backends.windows.WindowsBackend method) OfficialDesktopEnvBackend (class in dockyard_rl.experience.cua.backends.official) offload_after_refit() (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.interfaces.ColocatablePolicyInterface method) (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) offload_before_refit() (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.interfaces.ColocatablePolicyInterface method) (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) online_dpo_train() (in module dockyard_rl.algorithms.online_dpo) online_packing() (in module dockyard_rl.data.packing.algorithms) OnlineDPOConfig (class in dockyard_rl.algorithms.online_dpo) OnPolicyDistillationConfig (class in dockyard_rl.algorithms.opd) op (dockyard_rl.data_plane.observability.DataPlaneEvent attribute)

OPD OPDAdvantageEstimator (class in dockyard_rl.algorithms.advantage_estimator) OpenAIFormatDataset (class in dockyard_rl.data.datasets.response_datasets.oai_format_dataset) OpenMathInstruct2Dataset (class in dockyard_rl.data.datasets.response_datasets.openmathinstruct2) OptimizerState (class in dockyard_rl.utils.native_checkpoint) original_init (in module dockyard_rl.models.generation.vllm.quantization.fp8) original_run_engine_core (in module dockyard_rl.models.generation.vllm.quantization.fp8) ORPO orpo_lambda (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.loss.loss_functions.DPOLossConfig attribute) ORPOLossFn (class in dockyard_rl.algorithms.loss.loss_functions) osworld_data_processor() (in module dockyard_rl.experience.cua.datasets.osworld) OSWorldBackend (class in dockyard_rl.experience.cua.backends.interface) OSWorldDataset (class in dockyard_rl.experience.cua.datasets.osworld) OSWorldEnvironment (class in dockyard_rl.experience.cua.environment) output_ids (dockyard_rl.models.generation.interfaces.GenerationOutputSpec attribute) output_key (dockyard_rl.data.ResponseDatasetConfig attribute) output_limit (dockyard_rl.environments.gdpval_agentic_environment.GDPvalAgenticEnvConfig attribute) (dockyard_rl.environments.program_bench_environment.ProgramBenchEnvConfig attribute) (dockyard_rl.environments.terminal_bench_environment.TerminalBenchEnvConfig attribute) output_mount (dockyard_rl.sandbox.task_executor_client.TaskSpec attribute) output_splits (dockyard_rl.models.dtensor.moe.dispatch.AllToAllDispatchMetadata attribute) overlong_buffer_length (dockyard_rl.algorithms.reward_functions.RewardShapingConfig attribute) overlong_buffer_penalty (dockyard_rl.algorithms.reward_functions.RewardShapingConfig attribute) overlong_filtering (dockyard_rl.algorithms.grpo.GRPOConfig attribute)

P

Pack (in module dockyard_rl.data.packing.algorithms) pack_jagged_fields() (in module dockyard_rl.data_plane.codec) pack_per_token_field() (in module dockyard_rl.data_plane.codec) pack_sequences() (in module dockyard_rl.data.packing.algorithms) (in module dockyard_rl.models.huggingface.common) packed_broadcast_consumer() (in module dockyard_rl.utils.packed_tensor) packed_broadcast_producer() (in module dockyard_rl.utils.packed_tensor) PackedTensor (class in dockyard_rl.data.multimodal_utils) packing_efficiency (dockyard_rl.data.packing.metrics.PackingMetrics property) (dockyard_rl.data.packing.packing.PackingMetrics property) PackingMetrics (class in dockyard_rl.data.packing.metrics) (class in dockyard_rl.data.packing.packing) PackList (in module dockyard_rl.data.packing.algorithms) pad_distillation_val_batch() (in module dockyard_rl.algorithms.x_token.utils) padding_rate (dockyard_rl.data.packing.metrics.PackingMetrics property) (dockyard_rl.data.packing.packing.PackingMetrics property) pair_is_correct (dockyard_rl.algorithms.x_token.loss_utils.LocalizedAlignment attribute) (dockyard_rl.algorithms.x_token.token_aligner.AlignmentBatch attribute) pair_valid (dockyard_rl.algorithms.x_token.loss_utils.LocalizedAlignment attribute) (dockyard_rl.algorithms.x_token.token_aligner.AlignmentBatch attribute) pairs_to_preference_data() (in module dockyard_rl.algorithms.online_dpo) panel_aliases (dockyard_rl.algorithms.opd.TeacherPanelConfig attribute) ParallelismSpec (class in dockyard_rl.cluster.placement) parallelize_grouped_experts() (in module dockyard_rl.models.dtensor.moe.parallelize_moe) parallelize_router_gate() (in module dockyard_rl.models.dtensor.moe.parallelize_moe) PARALLIZE_FUNCTIONS (in module dockyard_rl.models.dtensor.parallelize) param_dtype (dockyard_rl.models.jax.config.JaxPolicyConfig attribute) param_named_shardings() (in module dockyard_rl.models.jax.sharding) parameters() (dockyard_rl.tool_protocol.registry.ToolRegistry method) parse_action() (in module dockyard_rl.experience.cua.actions.pyautogui_parse) parse_conversations() (in module dockyard_rl.data.datasets.response_datasets.oasst) parse_diff_paths() (in module dockyard_rl.rewards.diff_utils) parse_hermes_tool_calls() (in module dockyard_rl.tool_protocol.hermes) parse_hydra_overrides() (in module dockyard_rl.utils.config) parse_judge_score() (in module dockyard_rl.algorithms.online_dpo) parse_projection_file() (in module dockyard_rl.algorithms.x_token.loss_utils) parse_rubric() (in module dockyard_rl.rewards.gdpval_rubric) ParsedAction (class in dockyard_rl.experience.cua.actions.pyautogui_parse) ParsedToolCall (class in dockyard_rl.tool_protocol.hermes) partial_rotary_factor (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) partition_id (dockyard_rl.data_plane.interfaces.KVBatchMeta attribute) (dockyard_rl.data_plane.observability.DataPlaneEvent attribute) passthrough_prompt_response (dockyard_rl.data.chat_templates.COMMON_CHAT_TEMPLATES attribute) patch (dockyard_rl.sandbox.task_executor_client.TaskSpec attribute) patch_applied (dockyard_rl.sandbox.task_executor_client.TaskResult attribute) patch_diff (dockyard_rl.sandbox.task_executor_client.TaskResult attribute) patch_similarity (dockyard_rl.sandbox.task_executor_client.TaskResult attribute) path (dockyard_rl.utils.checkpoint.PretrainedCheckpointConfig attribute) path_needs_transpose() (in module dockyard_rl.models.jax.weights) PathLike (in module dockyard_rl.data.interfaces) (in module dockyard_rl.models.policy.lm_policy) (in module dockyard_rl.utils.checkpoint) pause() (dockyard_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollectorImpl method) PDB peak_bytes_outstanding (dockyard_rl.data_plane.observability.DataPlaneStats attribute) peft_config (dockyard_rl.utils.checkpoint.CheckpointingConfig attribute) penalty_mode (dockyard_rl.rewards.invalid_action.InvalidActionPenaltyConfig attribute) penalty_scale_at_step() (in module dockyard_rl.rewards.invalid_action) penalty_step_scale (dockyard_rl.rewards.invalid_action.InvalidActionPenaltyConfig attribute) per_expert_counts() (in module dockyard_rl.models.jax.moe.block) per_token_group_quant_fp8() (in module dockyard_rl.models.generation.vllm.quantization.fp8) permuted_indices (dockyard_rl.models.dtensor.moe.dispatch.AllToAllDispatchMetadata attribute) (dockyard_rl.models.jax.moe.alltoall.AllToAllDispatchMetadata attribute) PG pil_to_base64() (in module dockyard_rl.data.datasets.utils) pin_memory() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) pipeline_parallel_size (dockyard_rl.cluster.placement.ParallelismSpec attribute) (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) placement_strategy (dockyard_rl.cluster.fleet.FleetSpec attribute) PlacementConfigError Policy (class in dockyard_rl.models.policy.lm_policy) policy (dockyard_rl.algorithms.distillation.MasterConfig attribute) (dockyard_rl.algorithms.dpo.MasterConfig attribute) (dockyard_rl.algorithms.grpo.MasterConfig attribute) (dockyard_rl.algorithms.rm.MasterConfig attribute) (dockyard_rl.algorithms.sft.MasterConfig attribute) POLICY_WORKER_OVERRIDES (in module dockyard_rl.models.policy.utils) PolicyConfig (in module dockyard_rl.models.policy.interfaces) PolicyInterface (class in dockyard_rl.models.policy.interfaces) pop() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) pop_env_flags() (in module dockyard_rl.rewards.invalid_action) post_init() (dockyard_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) (dockyard_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) pow2_activation_scaling_factors (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) pow2_weight_scaling_factors (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) pp_size (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) PPO precision (dockyard_rl.algorithms.opd.TeacherResourceConfig attribute) (dockyard_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) (dockyard_rl.models.policy.teacher_worker_group.TeacherConfig attribute) prefer_verifiable (dockyard_rl.algorithms.online_dpo.OnlineDPOConfig attribute) preference_average_log_probs (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.loss.loss_functions.DPOLossConfig attribute) preference_collate_fn() (in module dockyard_rl.data.collate_fn) PREFERENCE_DATASET_REGISTRY (in module dockyard_rl.data.datasets.preference_datasets) preference_label (dockyard_rl.data.interfaces.KTODatumSpec attribute) preference_loss (dockyard_rl.algorithms.dpo.DPOValMetrics attribute)

PREFERENCE_LOSS_REGISTRY (in module dockyard_rl.algorithms.loss.loss_functions) preference_loss_weight (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.loss.loss_functions.DPOLossConfig attribute) preference_preprocessor() (in module dockyard_rl.data.processors) PreferenceDataset (class in dockyard_rl.data.datasets.preference_datasets.preference_dataset) PreferenceDatasetConfig (class in dockyard_rl.data) PreferenceDatumSpec (class in dockyard_rl.data.interfaces) PreferenceJudge (class in dockyard_rl.algorithms.online_dpo) PreferenceLossDataDict (class in dockyard_rl.algorithms.loss.loss_functions) PreferenceLossFn (class in dockyard_rl.algorithms.loss.loss_functions) prepare_for_generation() (dockyard_rl.models.generation.interfaces.GenerationInterface method) (dockyard_rl.models.generation.sglang.sglang_generation.SGLangGeneration method) (dockyard_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.interfaces.ColocatablePolicyInterface method) (dockyard_rl.models.policy.lm_policy.Policy method) prepare_for_lp_inference() (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.interfaces.ColocatablePolicyInterface method) (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) prepare_for_refit() (dockyard_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollectorImpl method) prepare_for_training() (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.interfaces.ColocatablePolicyInterface method) (dockyard_rl.models.policy.interfaces.PolicyInterface method) (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) prepare_gemma4_refit_info() (in module dockyard_rl.models.jax.models.gemma4_refit) prepare_loss_input() (in module dockyard_rl.algorithms.loss.utils) prepare_moe_refit_info() (in module dockyard_rl.models.jax.moe.refit) prepare_qwen3_next_refit_info() (in module dockyard_rl.models.jax.linear_attn.refit) prepare_refit_info() (dockyard_rl.models.generation.interfaces.GenerationInterface method) (dockyard_rl.models.generation.sglang.sglang_generation.SGLangGeneration method) (dockyard_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (dockyard_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (dockyard_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) (dockyard_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.interfaces.ColocatablePolicyInterface method) (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (dockyard_rl.models.policy.workers.dtensor_policy_worker_v2.DTensorPolicyWorkerV2Impl method) (in module dockyard_rl.models.jax.refit) prepare_segment_topology() (in module dockyard_rl.distributed.virtual_cluster) prepare_step() (dockyard_rl.models.policy.tq_policy.TQPolicy method) prepare_xtoken_cross_tokenizer_loss_input() (in module dockyard_rl.algorithms.x_token.loss_utils) preprocess_data() (dockyard_rl.environments.reward_model_environment.RewardModelEnvironment method) preprocessor (dockyard_rl.data.datasets.raw_dataset.RawDataset attribute) PreservingDataset (class in dockyard_rl.data.datasets.response_datasets.oai_format_dataset) pretrained_checkpoint (dockyard_rl.utils.checkpoint.CheckpointingConfig attribute) PretrainedCheckpointConfig (class in dockyard_rl.utils.checkpoint) prev_logprobs (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossDataDict attribute) principles (dockyard_rl.algorithms.rlaif.Constitution attribute) print_message_log_samples() (in module dockyard_rl.utils.logger) print_node_ip_and_gpu_id() (dockyard_rl.models.policy.lm_policy.Policy method) print_performance_metrics() (in module dockyard_rl.algorithms.utils) problem_key (dockyard_rl.data.LocalMathEvalDataConfig attribute) process_message_fragment() (dockyard_rl.data.datasets.response_datasets.general_conversation_dataset.GeneralConversationsJsonlDataset class method) process_weights_after_loading() (in module dockyard_rl.models.generation.vllm.quantization.fp8) process_weights_after_loading_kv() (in module dockyard_rl.models.generation.vllm.quantization.fp8) process_weights_after_loading_moe() (in module dockyard_rl.models.generation.vllm.quantization.fp8) processor (dockyard_rl.data.datasets.raw_dataset.RawDataset attribute) (dockyard_rl.data.ResponseDatasetConfig attribute) PROCESSOR_REGISTRY (in module dockyard_rl.data.processors) ProfilablePolicy (class in dockyard_rl.utils.nsys) program_bench_data_processor() (in module dockyard_rl.data.processors) ProgramBenchDataset (class in dockyard_rl.data.datasets.program_bench) ProgramBenchEnvConfig (class in dockyard_rl.environments.program_bench_environment) ProgramBenchEnvironment (class in dockyard_rl.environments.program_bench_environment) ProgramBenchReward (class in dockyard_rl.rewards.program_bench) project (dockyard_rl.utils.logger.SwanlabConfig attribute) (dockyard_rl.utils.logger.WandbConfig attribute) project_student_to_teacher_vocab() (in module dockyard_rl.algorithms.x_token.loss_utils) projection_matrix_paths (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) prompt_file (dockyard_rl.data.AIMEEvalDataConfig attribute) (dockyard_rl.data.GPQAEvalDataConfig attribute) (dockyard_rl.data.interfaces.TaskDataSpec attribute) (dockyard_rl.data.LocalMathEvalDataConfig attribute) (dockyard_rl.data.MathEvalDataConfig attribute) (dockyard_rl.data.MMAUEvalDataConfig attribute) (dockyard_rl.data.MMMLUEvalDataConfig attribute) (dockyard_rl.data.MMMLUProEvalDataConfig attribute) (dockyard_rl.data.PreferenceDatasetConfig attribute) (dockyard_rl.data.ResponseDatasetConfig attribute) prompt_key (dockyard_rl.data.PreferenceDatasetConfig attribute) prompt_text (dockyard_rl.algorithms.online_dpo.Candidate attribute) prompt_token_ids (dockyard_rl.algorithms.online_dpo.Candidate attribute) Provisioner (class in dockyard_rl.experience.cua.provisioner) pull_timeout (dockyard_rl.sandbox.task_executor_client.SessionStartSpec attribute) (dockyard_rl.sandbox.task_executor_client.TaskSpec attribute) put_samples() (dockyard_rl.data_plane.adapters.noop.NoOpDataPlaneClient method) (dockyard_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) (dockyard_rl.data_plane.interfaces.DataPlaneClient method) (dockyard_rl.data_plane.observability.MetricsDataPlaneClient method) PYAUTOGUI_PKGS_PREFIX (in module dockyard_rl.experience.cua.clients.guest_server) pyautogui_to_grammar() (in module dockyard_rl.experience.cua.actions.cc_transpile) pyautogui_to_macos() (in module dockyard_rl.experience.cua.actions.cc_transpile) pyautogui_to_windows() (in module dockyard_rl.experience.cua.actions.cc_transpile) pyautogui_to_xdotool() (in module dockyard_rl.experience.cua.actions.cc_transpile)

Q

quant_cfg (dockyard_rl.models.generation.vllm.config.VllmConfig attribute) quantization_ignored_layer_kws (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) Qwen3Attention (class in dockyard_rl.models.jax.models.qwen3) Qwen3Config (class in dockyard_rl.models.jax.models.qwen3) Qwen3DecoderLayer (class in dockyard_rl.models.jax.models.qwen3) Qwen3ForCausalLM (class in dockyard_rl.models.jax.models.qwen3) Qwen3MLP (class in dockyard_rl.models.jax.models.qwen3) Qwen3Model (class in dockyard_rl.models.jax.models.qwen3) Qwen3MoeConfig (class in dockyard_rl.models.jax.moe.qwen3_moe) Qwen3MoeDecoderLayer (class in dockyard_rl.models.jax.moe.qwen3_moe)

Qwen3MoeForCausalLM (class in dockyard_rl.models.jax.moe.qwen3_moe) Qwen3MoeModel (class in dockyard_rl.models.jax.moe.qwen3_moe) Qwen3NextAttention (class in dockyard_rl.models.jax.models.qwen3_next) Qwen3NextConfig (class in dockyard_rl.models.jax.models.qwen3_next) Qwen3NextDecoderLayer (class in dockyard_rl.models.jax.models.qwen3_next) Qwen3NextForCausalLM (class in dockyard_rl.models.jax.models.qwen3_next) Qwen3NextGatedDeltaNet (class in dockyard_rl.models.jax.linear_attn.gated_deltanet) Qwen3NextMLP (class in dockyard_rl.models.jax.models.qwen3_next) Qwen3NextModel (class in dockyard_rl.models.jax.models.qwen3_next) Qwen3NextSharedExpert (class in dockyard_rl.models.jax.models.qwen3_next)

R

R-DPO RAGEnvConfig (class in dockyard_rl.environments.retriever) RAGEnvironment (class in dockyard_rl.environments.retriever) RAGEnvironmentMetadata (class in dockyard_rl.environments.retriever) random_seed (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) ratio_clip_c (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) ratio_clip_max (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) ratio_clip_min (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) raw (dockyard_rl.experience.cua.actions.pyautogui_parse.ParsedAction attribute) RawDataset (class in dockyard_rl.data.datasets.raw_dataset) RayGpuMonitorLogger (class in dockyard_rl.utils.logger) RayVirtualCluster (class in dockyard_rl.distributed.virtual_cluster) RayWorkerBuilder (class in dockyard_rl.distributed.worker_groups) RayWorkerBuilder.IsolatedWorkerInitializer (class in dockyard_rl.distributed.worker_groups) RayWorkerGroup (class in dockyard_rl.distributed.worker_groups) RDMA RDPOLossFn (class in dockyard_rl.algorithms.loss.loss_functions) read_cell_value() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.utils) read_columns() (in module dockyard_rl.data_plane.column_io) read_config_value() (in module dockyard_rl.algorithms.config_access) read_from_dataplane() (dockyard_rl.models.policy.tq_policy.TQPolicy method) read_num_routed_experts() (in module dockyard_rl.models.dtensor.moe.config) real_quant (dockyard_rl.models.generation.vllm.config.VllmConfig attribute) real_quant_ignore (dockyard_rl.models.generation.vllm.config.VllmConfig attribute) reason (dockyard_rl.tool_protocol.registry.ToolUseAssessment attribute) reasoning_parser_plugin (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) rebuild_cuda_tensor_from_ipc() (in module dockyard_rl.models.policy.utils) rebuild_teacher_full_logits_cross_cluster() (in module dockyard_rl.algorithms.x_token.loss_utils) rebuild_teacher_full_logits_from_ipc() (in module dockyard_rl.algorithms.x_token.loss_utils) recompute_kv_cache_after_weight_updates (dockyard_rl.algorithms.grpo.AsyncGRPOConfig attribute) reconstruct_message_log() (in module dockyard_rl.data.llm_message_utils) recurrent_gated_delta_rule() (in module dockyard_rl.models.jax.linear_attn.delta_rule) recursive_merge_options() (in module dockyard_rl.distributed.worker_groups_utils) recv_refit_payload() (in module dockyard_rl.models.policy.utils) recv_shape (dockyard_rl.models.jax.moe.alltoall.AllToAllDispatchMetadata attribute) recv_sizes (dockyard_rl.models.jax.moe.alltoall.AllToAllDispatchMetadata attribute) reduce() (dockyard_rl.utils.timer.Timer method) ReduceTokensFn (in module dockyard_rl.models.dtensor.moe.load_balance) (in module dockyard_rl.models.jax.moe.load_balance) reference_dir (dockyard_rl.environments.gdpval_agentic_environment.GDPvalAgenticEnvConfig attribute) reference_free (dockyard_rl.algorithms.loss.loss_functions.DPOLossFn attribute) (dockyard_rl.algorithms.loss.loss_functions.ORPOLossFn attribute) (dockyard_rl.algorithms.loss.loss_functions.SimPOLossFn attribute) reference_logprobs (dockyard_rl.models.policy.interfaces.ReferenceLogprobOutputSpec attribute) reference_patch (dockyard_rl.sandbox.task_executor_client.TaskSpec attribute) reference_policy_kl_penalty (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) (dockyard_rl.algorithms.loss.loss_functions.DPOLossConfig attribute) reference_policy_kl_type (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) reference_policy_logprobs (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossDataDict attribute) (dockyard_rl.algorithms.loss.loss_functions.DPOLossDataDict attribute) ReferenceLogprobOutputSpec (class in dockyard_rl.models.policy.interfaces) refit_buffer_size_gb (dockyard_rl.models.generation.vllm.config.VllmConfig attribute) refit_socket_address() (in module dockyard_rl.models.policy.utils) refit_socket_dir() (in module dockyard_rl.models.policy.utils) REFIT_SOCKET_DIR_ENV (in module dockyard_rl.models.policy.utils) refit_tensor_dtype() (in module dockyard_rl.models.dtensor.moe.refit) regeneration_period (dockyard_rl.algorithms.online_dpo.OnlineDPOConfig attribute) register_env() (in module dockyard_rl.environments.utils) register_evaluator() (in module dockyard_rl.experience.cua.grading.registry) register_expert_bias_update_hook() (in module dockyard_rl.models.dtensor.moe.load_balance) register_omegaconf_resolvers() (in module dockyard_rl.utils.config) register_partition() (dockyard_rl.data_plane.adapters.noop.NoOpDataPlaneClient method) (dockyard_rl.data_plane.adapters.transfer_queue.TQDataPlaneClient method) (dockyard_rl.data_plane.interfaces.DataPlaneClient method) (dockyard_rl.data_plane.observability.MetricsDataPlaneClient method) register_processor() (in module dockyard_rl.data.processors) ReinforcePlusPlusAdvantageEstimator (class in dockyard_rl.algorithms.advantage_estimator) rejected_key (dockyard_rl.data.PreferenceDatasetConfig attribute) relativeTime_to_IntDay (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.misc) release() (dockyard_rl.experience.cua.provisioner.DockerPoolProvisioner method) (dockyard_rl.experience.cua.provisioner.Provisioner method) (dockyard_rl.experience.cua.provisioner.StaticPoolProvisioner method) remap_dataset_keys() (in module dockyard_rl.data.llm_message_utils) remove_boxed() (in module dockyard_rl.environments.dapo_math_verifier) remove_old_checkpoints() (dockyard_rl.utils.checkpoint.CheckpointManager method) REMOVED_EXPRESSIONS (in module dockyard_rl.environments.dapo_math_verifier) reorder_data() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) repeat_interleave() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) ReplayBuffer (class in dockyard_rl.algorithms.async_utils.replay_buffer) ReplayBufferImpl (class in dockyard_rl.algorithms.async_utils.replay_buffer) ReplayBufferProtocol (class in dockyard_rl.algorithms.async_utils.interfaces) REPLICATED_AXES (in module dockyard_rl.distributed.named_sharding) repo_url (dockyard_rl.sandbox.task_executor_client.TaskSpec attribute) report_device_id() (dockyard_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (dockyard_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) (dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) report_device_id_async() (dockyard_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) report_node_ip_and_gpu_id() (dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) required (dockyard_rl.rewards.gdpval_rubric.RubricItem attribute) requires_kv_scale_sync (dockyard_rl.models.generation.interfaces.GenerationInterface property) requires_prepadded_input (dockyard_rl.models.policy.teacher_scorer.TeacherScorer property) (dockyard_rl.models.policy.teacher_worker_group.TeacherWorkerGroup property) requires_unified_placement_group() (in module dockyard_rl.cluster.placement) reset() (dockyard_rl.data.packing.metrics.PackingMetrics method) (dockyard_rl.data.packing.packing.PackingMetrics method) (dockyard_rl.experience.cua.backends.control_center.ControlCenterBackend method) (dockyard_rl.experience.cua.backends.interface.OSWorldBackend method) (dockyard_rl.experience.cua.backends.macos.MacBackend method) (dockyard_rl.experience.cua.backends.official.OfficialDesktopEnvBackend method) (dockyard_rl.experience.cua.backends.windows.WindowsBackend method) (dockyard_rl.models.policy.interfaces.Timer method) (dockyard_rl.utils.timer.Timer method) reset_encoder_cache() (dockyard_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) reset_encoder_cache_async() (dockyard_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) reset_peak_memory_stats() (dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) reset_prefix_cache() (dockyard_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) reset_prefix_cache_async() (dockyard_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) resolve_device_capability() (in module dockyard_rl.utils.device_capability) resolve_device_query_backend() (in module dockyard_rl.utils.device_query) resolve_evaluator() (in module dockyard_rl.experience.cua.grading.registry) resolve_external_dataset_class() (in module dockyard_rl.data.datasets.response_datasets) resolve_generation_worker_cls() (in module dockyard_rl.models.generation.vllm.utils) resolve_invalid_action_cfg() (in module dockyard_rl.algorithms.grpo) resolve_judge() (in module dockyard_rl.rewards.gdpval_rubric)

resolve_model_class() (in module dockyard_rl.models.policy.utils) resolve_moe_parallelizer() (in module dockyard_rl.models.dtensor.moe.config) resolve_policy_worker_cls() (in module dockyard_rl.models.policy.utils) resolve_quant_cfg() (in module dockyard_rl.modelopt.utils) resolve_reference_aliases() (in module dockyard_rl.algorithms.opd) resolve_router_replay_enabled() (in module dockyard_rl.models.dtensor.moe.router_replay) resolve_sandbox_urls() (in module dockyard_rl.environments.multi_turn_session_env) resolve_structured_tool_use() (in module dockyard_rl.tool_protocol.protocol) resolve_structured_tool_use_cfg() (in module dockyard_rl.algorithms.grpo) resolve_to_image() (in module dockyard_rl.data.multimodal_utils) resolve_trainer_env() (in module dockyard_rl.models.jax.config) ResourceInsufficiencyError resources (dockyard_rl.algorithms.distillation.TeacherConfig attribute) (dockyard_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) respect_allow_internet (dockyard_rl.environments.terminal_bench_environment.TerminalBenchEnvConfig attribute) response_from_nested() (in module dockyard_rl.data_plane.codec) response_text (dockyard_rl.algorithms.online_dpo.Candidate attribute) response_token_counts() (in module dockyard_rl.algorithms.effort_reward) response_token_ids (dockyard_rl.algorithms.online_dpo.Candidate attribute) ResponseDataset (class in dockyard_rl.data.datasets.response_datasets.response_dataset) ResponseDatasetConfig (class in dockyard_rl.data) restore_checkpoint() (in module dockyard_rl.models.jax.checkpoint) restore_optimizer (dockyard_rl.utils.checkpoint.PretrainedCheckpointConfig attribute) result_files (dockyard_rl.sandbox.task_executor_client.FinishResult attribute) (dockyard_rl.sandbox.task_executor_client.TaskResult attribute) (dockyard_rl.sandbox.task_executor_client.TaskSpec attribute) resume() (dockyard_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollectorImpl method) resume_after_refit() (dockyard_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollectorImpl method) retrieve() (dockyard_rl.environments.retriever.BM25Retriever method) return_from_workers (dockyard_rl.distributed.worker_groups.MultiWorkerFuture attribute) return_model_config() (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) return_state_dict() (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) reverse_kl (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) revise() (in module dockyard_rl.algorithms.rlaif) revision_request (dockyard_rl.algorithms.rlaif.ConstitutionPrinciple attribute) reward (dockyard_rl.rewards.interfaces.RewardVerificationResult attribute) reward_floor (dockyard_rl.rewards.invalid_action.InvalidActionPenaltyConfig attribute) reward_functions (dockyard_rl.environments.vlm_environment.VLMEnvConfig attribute) reward_kind (dockyard_rl.environments.code_environment.CodeEnvConfig attribute) reward_mode (dockyard_rl.environments.code_environment.CodeEnvConfig attribute) (dockyard_rl.environments.gdpval_agentic_environment.GDPvalAgenticEnvConfig attribute) (dockyard_rl.environments.gdpval_environment.GDPvalEnvConfig attribute) (dockyard_rl.environments.program_bench_environment.ProgramBenchEnvConfig attribute) (dockyard_rl.environments.terminal_bench_environment.TerminalBenchEnvConfig attribute) REWARD_NAMES (dockyard_rl.environments.math_environment.HFMultiRewardVerifyWorker attribute) reward_scaling (dockyard_rl.algorithms.grpo.GRPOConfig attribute) reward_shaping (dockyard_rl.algorithms.grpo.GRPOConfig attribute) RewardFunction (class in dockyard_rl.rewards.interfaces) RewardModelEnvironment (class in dockyard_rl.environments.reward_model_environment) RewardModelEnvironmentConfig (class in dockyard_rl.environments.reward_model_environment) rewards (dockyard_rl.environments.interfaces.EnvironmentReturn attribute) rewards_chosen_mean (dockyard_rl.algorithms.dpo.DPOValMetrics attribute) (dockyard_rl.algorithms.rm.RMValMetrics attribute) rewards_rejected_mean (dockyard_rl.algorithms.dpo.DPOValMetrics attribute) (dockyard_rl.algorithms.rm.RMValMetrics attribute) RewardScalingConfig (class in dockyard_rl.algorithms.grpo) RewardShapingConfig (class in dockyard_rl.algorithms.reward_functions) RewardVerificationResult (class in dockyard_rl.rewards.interfaces) rl_collate_fn() (in module dockyard_rl.data.collate_fn) RLAIF RLOO RM rm (dockyard_rl.algorithms.rm.MasterConfig attribute) rm_train() (in module dockyard_rl.algorithms.rm) RMConfig (class in dockyard_rl.algorithms.rm) rms_norm_eps (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) (dockyard_rl.models.jax.models.qwen3.Qwen3Config attribute) (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) (dockyard_rl.models.jax.moe.qwen3_moe.Qwen3MoeConfig attribute) RMSaveState (class in dockyard_rl.algorithms.rm) RMSNorm (class in dockyard_rl.models.jax.layers) RMSNorm1p (class in dockyard_rl.models.jax.models.qwen3_next) RMSNormGated (class in dockyard_rl.models.jax.linear_attn.gated_deltanet) RMValMetrics (class in dockyard_rl.algorithms.rm) RoCE ROCR_VISIBLE_DEVICES (in module dockyard_rl.utils.device_query) role (dockyard_rl.cluster.fleet.FleetSpec attribute) rope_theta (dockyard_rl.models.jax.models.qwen3.Qwen3Config attribute) (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) (dockyard_rl.models.jax.moe.qwen3_moe.Qwen3MoeConfig attribute) rotary_dim (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig property) RotaryEmbedParallel (class in dockyard_rl.models.dtensor.parallelize) round_up() (in module dockyard_rl.data_plane.column_io) routed_expert_placements() (in module dockyard_rl.models.dtensor.moe.sharding) routed_experts (dockyard_rl.models.generation.interfaces.GenerationOutputSpec attribute) routed_experts_empty() (in module dockyard_rl.models.generation.vllm.router_capture) ROUTED_EXPERTS_FALLBACK_DTYPE (in module dockyard_rl.models.generation.vllm.router_capture) routed_experts_from_vllm_output() (in module dockyard_rl.models.generation.vllm.router_capture) Router replay (R3) router_gate_placements() (in module dockyard_rl.models.dtensor.moe.sharding) router_replay_context() (in module dockyard_rl.models.dtensor.moe.router_replay) routing_diversity (dockyard_rl.algorithms.grpo.GRPOConfig attribute) routing_diversity_enabled() (in module dockyard_rl.algorithms.routing_diversity) routing_diversity_from_message_log() (in module dockyard_rl.algorithms.routing_diversity) ROUTING_DIVERSITY_KEY (in module dockyard_rl.algorithms.routing_diversity) routing_diversity_metrics() (in module dockyard_rl.algorithms.routing_diversity) routing_map_and_counts() (in module dockyard_rl.models.dtensor.moe.block) RoutingDiversityConfig (class in dockyard_rl.algorithms.routing_diversity) RubricItem (class in dockyard_rl.rewards.gdpval_rubric) Rule (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.getters.misc) run() (dockyard_rl.models.generation.sglang.utils.AsyncLoopThread method) run_all_workers_multiple_data() (dockyard_rl.distributed.worker_groups.RayWorkerGroup method) (dockyard_rl.models.policy.lm_policy.Policy method) run_all_workers_sharded_data() (dockyard_rl.distributed.worker_groups.RayWorkerGroup method) run_all_workers_single_data() (dockyard_rl.distributed.worker_groups.RayWorkerGroup method) (dockyard_rl.models.policy.lm_policy.Policy method) run_async_cua_rollout() (in module dockyard_rl.experience.cua.rollout) run_async_multi_turn_rollout() (in module dockyard_rl.experience.rollouts) run_id (dockyard_rl.utils.logger.MLflowConfig attribute) run_multi_turn_rollout() (in module dockyard_rl.experience.rollouts) run_name (dockyard_rl.utils.logger.MLflowConfig attribute) run_online_dpo_iteration() (in module dockyard_rl.algorithms.online_dpo) run_sample_multi_turn_rollout() (in module dockyard_rl.experience.rollouts) run_single_worker_single_data() (dockyard_rl.distributed.worker_groups.RayWorkerGroup method) run_sqlite3() (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.general) run_task() (in module dockyard_rl.sandbox.task_executor_client)

S

s_end (dockyard_rl.algorithms.x_token.token_aligner.AlignmentPair attribute) s_start (dockyard_rl.algorithms.x_token.token_aligner.AlignmentPair attribute) s_tokens (dockyard_rl.algorithms.x_token.token_aligner.AlignmentPair attribute) sample() (dockyard_rl.algorithms.async_utils.interfaces.ReplayBufferProtocol method) (dockyard_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) sample_ids (dockyard_rl.data_plane.interfaces.KVBatchMeta attribute) sample_locus_penalty() (in module dockyard_rl.rewards.invalid_action) sample_mask (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossDataDict attribute) (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossDataDict attribute) (dockyard_rl.algorithms.loss.loss_functions.DistillationLossDataDict attribute) (dockyard_rl.algorithms.loss.loss_functions.DPOLossDataDict attribute) (dockyard_rl.algorithms.loss.loss_functions.DraftCrossEntropyLossDataDict attribute) (dockyard_rl.algorithms.loss.loss_functions.PreferenceLossDataDict attribute) (dockyard_rl.algorithms.x_token.loss_utils.LocalizedAlignment attribute) SAMPLE_MASK (in module dockyard_rl.data_plane.schema) sample_routing_diversity() (in module dockyard_rl.algorithms.routing_diversity) sampling_backend (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) sandbox_urls (dockyard_rl.environments.code_environment.CodeEnvConfig attribute) (dockyard_rl.environments.gdpval_agentic_environment.GDPvalAgenticEnvConfig attribute) (dockyard_rl.environments.program_bench_environment.ProgramBenchEnvConfig attribute) (dockyard_rl.environments.terminal_bench_environment.TerminalBenchEnvConfig attribute) save_checkpoint() (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.interfaces.PolicyInterface method) (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) (in module dockyard_rl.models.jax.checkpoint) (in module dockyard_rl.utils.native_checkpoint) save_consolidated (dockyard_rl.utils.checkpoint.CheckpointingConfig attribute) save_optimizer (dockyard_rl.utils.checkpoint.CheckpointingConfig attribute) save_period (dockyard_rl.utils.checkpoint.CheckpointingConfig attribute) scale_rewards() (in module dockyard_rl.algorithms.grpo) Schedule (in module dockyard_rl.models.jax.optimizer) schedule_conservativeness (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) schedule_policy (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) SchemaError score (dockyard_rl.algorithms.online_dpo.Candidate attribute) score() (dockyard_rl.algorithms.online_dpo.PreferenceJudge method) (dockyard_rl.environments.code_environment.SWEScoringWorker method) (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.teacher_scorer.TeacherScorer method) (dockyard_rl.models.policy.teacher_worker_group.TeacherWorkerGroup method) (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) score_margin (dockyard_rl.algorithms.online_dpo.OnlineDPOConfig attribute) ScoreFunc (in module dockyard_rl.models.dtensor.moe.router) (in module dockyard_rl.models.jax.moe.router) ScoreOutputSpec (class in dockyard_rl.models.policy.interfaces) scores (dockyard_rl.models.policy.interfaces.ScoreOutputSpec attribute) scoring_phase_bytes (dockyard_rl.algorithms.distillation_memory.MemoryReport property) screenshot (dockyard_rl.experience.cua.backends.interface.Observation attribute) screenshot_judge() (in module dockyard_rl.experience.cua.grading.builtins) screenshot_to_data_url() (in module dockyard_rl.experience.cua.vision) screenshot_to_image() (in module dockyard_rl.experience.cua.vision) seed (dockyard_rl.algorithms.distillation.DistillationConfig attribute) (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.grpo.GRPOConfig attribute) (dockyard_rl.algorithms.rm.RMConfig attribute) (dockyard_rl.algorithms.sft.SFTConfig attribute) (dockyard_rl.data.ResponseDatasetConfig attribute) seen_params (dockyard_rl.models.generation.vllm.quantization.fp8.FP8State attribute) segment_size (dockyard_rl.cluster.fleet.FleetSpec attribute) select() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) select_colocation_schedule() (in module dockyard_rl.algorithms.distillation_memory) select_escalation_rows() (in module dockyard_rl.algorithms.teacher_panel) select_indices() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) select_rows() (in module dockyard_rl.models.policy.teacher_scorer) select_segment_nodes() (in module dockyard_rl.distributed.virtual_cluster) select_teacher_topk_indices() (in module dockyard_rl.algorithms.x_token.loss_utils) send_sizes (dockyard_rl.models.jax.moe.alltoall.AllToAllDispatchMetadata attribute) seq_logprob_error_threshold (dockyard_rl.algorithms.grpo.GRPOConfig attribute) sequence_length_histogram (dockyard_rl.data.packing.metrics.PackingMetrics attribute) (dockyard_rl.data.packing.packing.PackingMetrics attribute) sequence_length_pad_multiple (dockyard_rl.distributed.batched_data_dict.SequencePackingArgs attribute) sequence_length_round (dockyard_rl.distributed.batched_data_dict.DynamicBatchingArgs attribute) sequence_lengths (dockyard_rl.data_plane.interfaces.KVBatchMeta attribute) SEQUENCE_LEVEL (dockyard_rl.algorithms.loss.interfaces.LossType attribute) sequence_level_importance_ratios (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) sequence_packing (dockyard_rl.environments.reward_model_environment.RewardModelEnvironmentConfig attribute) sequence_pad_multiple (dockyard_rl.models.policy.teacher_scorer.TeacherScorer property) (dockyard_rl.models.policy.teacher_worker_group.TeacherWorkerGroup property) SequencePackingArgs (class in dockyard_rl.distributed.batched_data_dict) SequencePackingFusionLossWrapper (class in dockyard_rl.algorithms.loss.wrapper) SequencePackingLossWrapper (class in dockyard_rl.algorithms.loss.wrapper) sequences_per_datum (dockyard_rl.algorithms.loss.loss_functions.DPOLossFn attribute) (dockyard_rl.algorithms.loss.loss_functions.KTOLossFn attribute) serialize() (dockyard_rl.models.generation.sglang.sglang_copied_utils.MultiprocessingSerializer static method) server_port (dockyard_rl.experience.cua.vendor.desktop_env_eval.grading_env.GradingEnv attribute) session_exec() (in module dockyard_rl.sandbox.task_executor_client) session_export() (in module dockyard_rl.sandbox.task_executor_client) session_finish() (in module dockyard_rl.sandbox.task_executor_client) session_id (dockyard_rl.sandbox.task_executor_client.ExecResult attribute) (dockyard_rl.sandbox.task_executor_client.FinishResult attribute) SESSION_TOOLS (in module dockyard_rl.tool_protocol.registry) SessionStartSpec (class in dockyard_rl.sandbox.task_executor_client) set_device_query_backend() (in module dockyard_rl.utils.device_query) set_processor() (dockyard_rl.data.datasets.raw_dataset.RawDataset method) set_records() (dockyard_rl.data.dataloader.MultipleDataloaderWrapper method) set_seed() (in module dockyard_rl.algorithms.utils) set_task_spec() (dockyard_rl.data.datasets.raw_dataset.RawDataset method) set_weight_version() (dockyard_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollectorImpl method) setup() (dockyard_rl.experience.cua.clients.guest_server.SetupController method) (in module dockyard_rl.algorithms.distillation) (in module dockyard_rl.algorithms.dpo) (in module dockyard_rl.algorithms.grpo) (in module dockyard_rl.algorithms.rm) (in module dockyard_rl.algorithms.sft) setup_data_plane() (dockyard_rl.data_plane.worker_mixin.TQWorkerMixin method) setup_preference_data() (in module dockyard_rl.data.utils) setup_response_data() (in module dockyard_rl.data.utils) SetupController (class in dockyard_rl.experience.cua.clients.guest_server) severity (dockyard_rl.rewards.invalid_action.Violation attribute) severity_weights (dockyard_rl.rewards.invalid_action.InvalidActionPenaltyConfig attribute) SFT sft (dockyard_rl.algorithms.sft.MasterConfig attribute) sft_average_log_probs (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.loss.loss_functions.DPOLossConfig attribute) sft_loss (dockyard_rl.algorithms.dpo.DPOValMetrics attribute) sft_loss_weight (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.loss.loss_functions.DPOLossConfig attribute) sft_processor() (in module dockyard_rl.data.processors) sft_train() (in module dockyard_rl.algorithms.sft) SFTConfig (class in dockyard_rl.algorithms.sft) SFTSaveState (class in dockyard_rl.algorithms.sft) SGLANG_BACKEND (in module dockyard_rl.models.generation.constants) sglang_cfg (dockyard_rl.models.generation.sglang.config.SGLangConfig attribute) sglang_kwargs (dockyard_rl.models.generation.sglang.config.SGLangConfig attribute) SGLangConfig (class in dockyard_rl.models.generation.sglang.config) SGLangGeneration (class in dockyard_rl.models.generation.sglang.sglang_generation) SGLangGenerationWorker (class in dockyard_rl.models.generation.sglang.sglang_worker) SglangSpecificArgs (class in dockyard_rl.models.generation.sglang.config) shape (dockyard_rl.distributed.named_sharding.NamedSharding property) shard_by_batch_size() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) shard_meta_for_dp() (in module dockyard_rl.data_plane.preshard) shard_model() (in module dockyard_rl.models.jax.sharding) Sharding (in module dockyard_rl.models.jax.layers) shared_expert_intermediate_size (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) SHEET (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.table) show_time_cost (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) shuffle (dockyard_rl.data.DataConfig attribute) shuffle_packing() (in module dockyard_rl.data.packing.algorithms) shutdown() (dockyard_rl.distributed.virtual_cluster.RayVirtualCluster method) (dockyard_rl.distributed.worker_groups.RayWorkerGroup method) (dockyard_rl.environments.code_environment.CodeEnvironment method) (dockyard_rl.environments.code_jaccard_environment.CodeJaccardEnvironment method) (dockyard_rl.environments.gdpval_environment.GDPvalEnvironment method) (dockyard_rl.environments.hle_environment.HLEEnvironment method) (dockyard_rl.environments.math_environment.BaseMathEnvironment method) (dockyard_rl.environments.multi_turn_session_env.MultiTurnSessionEnvironment method) (dockyard_rl.environments.retriever.RAGEnvironment method) (dockyard_rl.environments.reward_model_environment.RewardModelEnvironment method) (dockyard_rl.environments.vlm_environment.VLMEnvironment method) (dockyard_rl.experience.cua.environment.OSWorldEnvironment method) (dockyard_rl.experience.cua.provisioner.DockerPoolProvisioner method) (dockyard_rl.experience.cua.provisioner.Provisioner method) (dockyard_rl.models.generation.sglang.sglang_generation.SGLangGeneration method) (dockyard_rl.models.generation.sglang.sglang_worker.SGLangGenerationWorker method) (dockyard_rl.models.generation.sglang.utils.AsyncLoopThread method) (dockyard_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (dockyard_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) (dockyard_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.interfaces.PolicyInterface method) (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.teacher_scorer.TeacherScorer method) (dockyard_rl.models.policy.teacher_worker_group.TeacherWorkerGroup method) (dockyard_rl.models.policy.tq_policy.TQPolicy method) (dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) (dockyard_rl.weight_sync.collective_weight_synchronizer.CollectiveWeightSynchronizer method) (dockyard_rl.weight_sync.http_weight_synchronizer.HTTPWeightSynchronizer method) (dockyard_rl.weight_sync.interfaces.WeightSynchronizer method) (dockyard_rl.weight_sync.ipc_weight_synchronizer.IPCWeightSynchronizer method) simple_role_header (dockyard_rl.data.chat_templates.COMMON_CHAT_TEMPLATES attribute) SimPO simpo_gamma (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.loss.loss_functions.DPOLossConfig attribute) SimPOLossFn (class in dockyard_rl.algorithms.loss.loss_functions) size (dockyard_rl.data_plane.interfaces.KVBatchMeta property) (dockyard_rl.distributed.batched_data_dict.BatchedDataDict property) (dockyard_rl.distributed.named_sharding.NamedSharding property) size() (dockyard_rl.algorithms.async_utils.interfaces.ReplayBufferProtocol method) (dockyard_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) skip_reference_policy_logprobs_calculation (dockyard_rl.algorithms.grpo.GRPOConfig attribute) skip_server_warmup (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) skip_tokenizer_init (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute)

sleep() (dockyard_rl.models.generation.sglang.sglang_worker.SGLangGenerationWorker method) (dockyard_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) sleep_async() (dockyard_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) slice() (dockyard_rl.data.multimodal_utils.PackedTensor method) (dockyard_rl.data_plane.interfaces.KVBatchMeta method) slice_sparse_projection_cols() (in module dockyard_rl.algorithms.x_token.loss_utils) slice_sparse_projection_rows() (in module dockyard_rl.algorithms.x_token.loss_utils) SlicedDataDict (class in dockyard_rl.distributed.batched_data_dict) sliding_rope_theta (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) sliding_rope_type (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) sliding_window (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) sm (dockyard_rl.utils.device_capability.DeviceCapability property) SM_BLOCK_FP8 (in module dockyard_rl.utils.device_capability) SM_FP8 (in module dockyard_rl.utils.device_capability) SM_NATIVE_BF16 (in module dockyard_rl.utils.device_capability) SM_PREFIX_CACHING (in module dockyard_rl.utils.device_capability) snapshot() (dockyard_rl.data_plane.observability.MetricsDataPlaneClient method) (dockyard_rl.experience.cua.vision.TheEyeClient method) snapshot_start_of_stage() (dockyard_rl.utils.memory_tracker.MemoryTracker method) solution_key (dockyard_rl.data.LocalMathEvalDataConfig attribute) source_max (dockyard_rl.algorithms.grpo.RewardScalingConfig attribute) source_min (dockyard_rl.algorithms.grpo.RewardScalingConfig attribute) split (dockyard_rl.data.LocalMathEvalDataConfig attribute) (dockyard_rl.data.MMAUEvalDataConfig attribute) (dockyard_rl.data.PreferenceDatasetConfig attribute) (dockyard_rl.data.ResponseDatasetConfig attribute) split() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) split_for_dp() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) split_microbatches() (in module dockyard_rl.models.jax.train_step) split_output_tensor() (dockyard_rl.algorithms.loss.loss_functions.PreferenceLossFn method) split_train_validation() (dockyard_rl.data.datasets.raw_dataset.RawDataset method) split_validation_size (dockyard_rl.data.ResponseDatasetConfig attribute) SquadDataset (class in dockyard_rl.data.datasets.response_datasets.squad) stack_or_nest() (in module dockyard_rl.data_plane.codec) stack_routed_experts() (in module dockyard_rl.models.generation.vllm.router_capture) stage (dockyard_rl.utils.memory_tracker.MemoryTrackerDataPoint attribute) stamp_tags() (dockyard_rl.data_plane.interfaces.KVBatchMeta method) start() (dockyard_rl.experience.cua.backends.control_center.ControlCenterBackend method) (dockyard_rl.experience.cua.backends.interface.OSWorldBackend method) (dockyard_rl.experience.cua.backends.macos.MacBackend method) (dockyard_rl.experience.cua.backends.official.OfficialDesktopEnvBackend method) (dockyard_rl.experience.cua.backends.windows.WindowsBackend method) (dockyard_rl.experience.cua.provisioner.DockerPoolProvisioner method) (dockyard_rl.utils.logger.RayGpuMonitorLogger method) (dockyard_rl.utils.timer.Timer method) start_collection() (dockyard_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollectorImpl method) start_gpu_profiling() (dockyard_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (dockyard_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker method) (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) (dockyard_rl.utils.nsys.ProfilablePolicy method) start_iterations() (dockyard_rl.utils.timer.TimeoutChecker method) start_session() (in module dockyard_rl.sandbox.task_executor_client) startup_command (dockyard_rl.sandbox.task_executor_client.SessionStartSpec attribute) state_dict() (dockyard_rl.algorithms.async_utils.replay_buffer.ReplayBufferImpl method) (dockyard_rl.utils.native_checkpoint.ModelState method) (dockyard_rl.utils.native_checkpoint.OptimizerState method) StatelessProcessGroup (class in dockyard_rl.distributed.stateless_process_group) StaticPoolProvisioner (class in dockyard_rl.experience.cua.provisioner) status (dockyard_rl.data_plane.observability.DataPlaneEvent attribute) (dockyard_rl.rewards.interfaces.RewardVerificationResult attribute) (dockyard_rl.sandbox.task_executor_client.TaskResult attribute) stderr (dockyard_rl.sandbox.task_executor_client.ExecResult attribute) (dockyard_rl.sandbox.task_executor_client.FinishResult attribute) (dockyard_rl.sandbox.task_executor_client.TaskResult attribute) stdout (dockyard_rl.sandbox.task_executor_client.ExecResult attribute) (dockyard_rl.sandbox.task_executor_client.FinishResult attribute) (dockyard_rl.sandbox.task_executor_client.TaskResult attribute) step (dockyard_rl.algorithms.distillation.DistillationSaveState attribute) (dockyard_rl.algorithms.dpo.DPOSaveState attribute) (dockyard_rl.algorithms.rm.RMSaveState attribute) (dockyard_rl.algorithms.sft.SFTSaveState attribute) (dockyard_rl.utils.logger.GpuMetricSnapshot attribute) step() (dockyard_rl.environments.code_environment.CodeEnvironment method) (dockyard_rl.environments.code_jaccard_environment.CodeJaccardEnvironment method) (dockyard_rl.environments.gdpval_environment.GDPvalEnvironment method) (dockyard_rl.environments.hle_environment.HLEEnvironment method) (dockyard_rl.environments.interfaces.EnvironmentInterface method) (dockyard_rl.environments.math_environment.MathEnvironment method) (dockyard_rl.environments.math_environment.MathMultiRewardEnvironment method) (dockyard_rl.environments.multi_turn_session_env.MultiTurnSessionEnvironment method) (dockyard_rl.environments.retriever.RAGEnvironment method) (dockyard_rl.environments.reward_model_environment.RewardModelEnvironment method) (dockyard_rl.environments.vlm_environment.VLMEnvironment method) (dockyard_rl.experience.cua.environment.OSWorldEnvironment method) stop() (dockyard_rl.algorithms.async_utils.trajectory_collector.AsyncTrajectoryCollectorImpl method) (dockyard_rl.experience.cua.backends.control_center.ControlCenterBackend method) (dockyard_rl.experience.cua.backends.interface.OSWorldBackend method) (dockyard_rl.experience.cua.backends.macos.MacBackend method) (dockyard_rl.experience.cua.backends.official.OfficialDesktopEnvBackend method) (dockyard_rl.experience.cua.backends.windows.WindowsBackend method) (dockyard_rl.utils.logger.RayGpuMonitorLogger method) (dockyard_rl.utils.timer.Timer method) stop_gpu_profiling() (dockyard_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (dockyard_rl.models.generation.vllm.vllm_worker.BaseVllmGenerationWorker method) (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.workers.base_policy_worker.AbstractPolicyWorker method) (dockyard_rl.utils.nsys.ProfilablePolicy method) stop_properly_penalty_coef (dockyard_rl.algorithms.reward_functions.RewardShapingConfig attribute) stop_strings (dockyard_rl.data.interfaces.DatumSpec attribute) (dockyard_rl.environments.code_jaccard_environment.CodeJaccardEnvConfig attribute) (dockyard_rl.environments.math_environment.MathEnvConfig attribute) (dockyard_rl.environments.retriever.RAGEnvConfig attribute) STOP_STRINGS (dockyard_rl.environments.retriever.RAGEnvironment attribute) stop_strings (dockyard_rl.environments.vlm_environment.VLMEnvConfig attribute) (dockyard_rl.models.generation.interfaces.GenerationConfig attribute) (dockyard_rl.models.generation.interfaces.GenerationDatumSpec attribute) stop_token_ids (dockyard_rl.models.generation.interfaces.GenerationConfig attribute) storage_capacity (dockyard_rl.data_plane.interfaces.DataPlaneConfig attribute) STRATEGY_PACK (in module dockyard_rl.cluster.fleet) STRATEGY_SPREAD (in module dockyard_rl.cluster.fleet) STRATEGY_STRICT_PACK (in module dockyard_rl.cluster.fleet) stream_weights_via_http() (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) stream_weights_via_http_impl() (in module dockyard_rl.models.policy.utils) stream_weights_via_ipc_zmq() (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) stream_weights_via_ipc_zmq_impl() (in module dockyard_rl.models.policy.utils) strict_agent_name_match (dockyard_rl.algorithms.opd.OnPolicyDistillationConfig attribute) structural_tag (dockyard_rl.models.generation.interfaces.StructuredSpec attribute) structural_token_mask (dockyard_rl.models.generation.interfaces.GenerationOutputSpec attribute) structured_spec (dockyard_rl.models.generation.interfaces.GenerationDatumSpec attribute) structured_tool_use (dockyard_rl.algorithms.grpo.GRPOConfig attribute) (dockyard_rl.data.interfaces.TaskDataSpec attribute) STRUCTURED_TOOL_USE_KEY (in module dockyard_rl.tool_protocol.protocol) structured_tools (dockyard_rl.environments.code_environment.CodeEnvConfig attribute) StructuredSpec (class in dockyard_rl.models.generation.interfaces) StructuredToolUseConfig (class in dockyard_rl.tool_protocol.protocol) student_activation_bytes (dockyard_rl.algorithms.distillation_memory.MemoryReport attribute) student_chunk_id (dockyard_rl.algorithms.x_token.loss_utils.LocalizedAlignment attribute) (dockyard_rl.algorithms.x_token.token_aligner.AlignmentBatch attribute) student_exact_partition_mask (dockyard_rl.algorithms.x_token.token_aligner.AlignmentBatch attribute) student_grad_bytes (dockyard_rl.algorithms.distillation_memory.MemoryReport attribute) student_input_ids (dockyard_rl.algorithms.x_token.loss_utils.LocalizedAlignment attribute) student_logits (dockyard_rl.algorithms.loss.loss_functions.DraftCrossEntropyLossDataDict attribute) student_next_token_ce() (in module dockyard_rl.algorithms.x_token.loss_utils) student_optimizer_bytes (dockyard_rl.algorithms.distillation_memory.MemoryReport attribute) student_token_mask (dockyard_rl.algorithms.x_token.loss_utils.LocalizedAlignment attribute) student_vocab_indices (dockyard_rl.algorithms.loss.loss_functions.DraftCrossEntropyLossDataDict attribute) student_vocab_size (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) student_weight_bytes (dockyard_rl.algorithms.distillation_memory.MemoryReport attribute) submit_task() (in module dockyard_rl.sandbox.task_executor_client) subset (dockyard_rl.data.PreferenceDatasetConfig attribute) (dockyard_rl.data.ResponseDatasetConfig attribute) subset() (dockyard_rl.data_plane.interfaces.KVBatchMeta method) SUBSTITUTIONS (in module dockyard_rl.environments.dapo_math_verifier) sum_weights_metric (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) SUPPORTED_ESTIMATORS (in module dockyard_rl.algorithms.effort_reward) SUPPORTED_TRAINERS (in module dockyard_rl.algorithms.effort_reward) (in module dockyard_rl.algorithms.rollout_allocation) (in module dockyard_rl.algorithms.routing_diversity) (in module dockyard_rl.algorithms.teacher_panel) supports_block_fp8 (dockyard_rl.utils.device_capability.DeviceCapability property) supports_fp8 (dockyard_rl.utils.device_capability.DeviceCapability property) supports_native_bf16 (dockyard_rl.utils.device_capability.DeviceCapability property) supports_prefix_caching (dockyard_rl.utils.device_capability.DeviceCapability property) suppress_ray_visible_devices() (in module dockyard_rl.utils.device_query) surpress_user_warnings() (in module dockyard_rl.algorithms.utils) swanlab (dockyard_rl.utils.logger.LoggerConfig attribute) swanlab_enabled (dockyard_rl.utils.logger.LoggerConfig attribute) SwanlabConfig (class in dockyard_rl.utils.logger) SwanlabLogger (class in dockyard_rl.utils.logger) SWE-bench swe_bench_data_processor() (in module dockyard_rl.data.processors) swe_bench_pro_data_processor() (in module dockyard_rl.data.processors) SWEBenchDataset (class in dockyard_rl.data.datasets.swe_bench) SWEBenchProDataset (class in dockyard_rl.data.datasets.swe_bench_pro) SWEBenchProReward (class in dockyard_rl.rewards.swe_bench_pro) SWEScoringWorker (class in dockyard_rl.environments.code_environment) sync_weights() (dockyard_rl.weight_sync.collective_weight_synchronizer.CollectiveWeightSynchronizer method) (dockyard_rl.weight_sync.http_weight_synchronizer.HTTPWeightSynchronizer method) (dockyard_rl.weight_sync.interfaces.WeightSynchronizer method) (dockyard_rl.weight_sync.ipc_weight_synchronizer.IPCWeightSynchronizer method) SYSTEM_PROMPT (in module dockyard_rl.data.datasets.response_datasets.oasst) system_prompt_file (dockyard_rl.data.AIMEEvalDataConfig attribute) (dockyard_rl.data.GPQAEvalDataConfig attribute) (dockyard_rl.data.interfaces.TaskDataSpec attribute) (dockyard_rl.data.LocalMathEvalDataConfig attribute) (dockyard_rl.data.MathEvalDataConfig attribute) (dockyard_rl.data.MMAUEvalDataConfig attribute) (dockyard_rl.data.MMMLUEvalDataConfig attribute) (dockyard_rl.data.MMMLUProEvalDataConfig attribute) (dockyard_rl.data.PreferenceDatasetConfig attribute) (dockyard_rl.data.ResponseDatasetConfig attribute)

T

T (in module dockyard_rl.models.dtensor.moe.refit) (in module dockyard_rl.models.jax.moe.refit) t_end (dockyard_rl.algorithms.x_token.token_aligner.AlignmentPair attribute) t_start (dockyard_rl.algorithms.x_token.token_aligner.AlignmentPair attribute) t_tokens (dockyard_rl.algorithms.x_token.token_aligner.AlignmentPair attribute) tags (dockyard_rl.data_plane.interfaces.KVBatchMeta attribute) target_max (dockyard_rl.algorithms.grpo.RewardScalingConfig attribute) target_min (dockyard_rl.algorithms.grpo.RewardScalingConfig attribute) task_id (dockyard_rl.sandbox.task_executor_client.TaskResult attribute) task_name (dockyard_rl.data.datasets.preference_datasets.binary_preference_dataset.BinaryPreferenceDataset attribute) (dockyard_rl.data.datasets.preference_datasets.helpsteer3.HelpSteer3PreferenceDataset attribute) (dockyard_rl.data.datasets.preference_datasets.preference_dataset.PreferenceDataset attribute) (dockyard_rl.data.datasets.preference_datasets.tulu3.Tulu3PreferenceDataset attribute) (dockyard_rl.data.datasets.response_datasets.avqa.AVQADataset attribute) (dockyard_rl.data.datasets.response_datasets.daily_omni.DailyOmniDataset attribute) (dockyard_rl.data.datasets.response_datasets.general_conversation_dataset.GeneralConversationsJsonlDataset attribute) (dockyard_rl.data.interfaces.DatumSpec attribute) (dockyard_rl.data.interfaces.KTODatumSpec attribute) (dockyard_rl.data.interfaces.TaskDataSpec attribute) (dockyard_rl.data_plane.interfaces.KVBatchMeta attribute) task_spec (dockyard_rl.data.datasets.raw_dataset.RawDataset attribute) task_timeout (dockyard_rl.environments.code_environment.CodeEnvConfig attribute) TaskConfigError TaskDataPreProcessFnCallable (class in dockyard_rl.data.interfaces) TaskDataProcessFnCallable (class in dockyard_rl.data.interfaces) TaskDataSpec (class in dockyard_rl.data.interfaces) TaskExecutorError TaskResult (class in dockyard_rl.sandbox.task_executor_client) TaskSpec (class in dockyard_rl.sandbox.task_executor_client) teacher_activation_bytes (dockyard_rl.algorithms.distillation_memory.MemoryReport attribute) teacher_chunk_id (dockyard_rl.algorithms.x_token.loss_utils.LocalizedAlignment attribute) (dockyard_rl.algorithms.x_token.token_aligner.AlignmentBatch attribute) teacher_exact_partition_mask (dockyard_rl.algorithms.x_token.token_aligner.AlignmentBatch attribute) teacher_gold_loss (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) teacher_inference_micro_batch_size (dockyard_rl.algorithms.distillation.DistillationConfig attribute) teacher_logits (dockyard_rl.algorithms.loss.loss_functions.DraftCrossEntropyLossDataDict attribute) teacher_model_by_agent_name (dockyard_rl.algorithms.opd.OnPolicyDistillationConfig attribute) teacher_overrides (dockyard_rl.algorithms.opd.NonColocatedTeachersConfig attribute) teacher_panel (dockyard_rl.algorithms.opd.OnPolicyDistillationConfig attribute) teacher_panel_cfg() (in module dockyard_rl.algorithms.teacher_panel) teacher_panel_enabled() (in module dockyard_rl.algorithms.teacher_panel) teacher_panel_metrics() (in module dockyard_rl.algorithms.teacher_panel) teacher_policy (dockyard_rl.algorithms.distillation.MasterConfig attribute) teacher_seq_pad_multiple() (in module dockyard_rl.algorithms.opd) teacher_topk_indices (dockyard_rl.algorithms.loss.loss_functions.DistillationLossDataDict attribute) teacher_topk_logits (dockyard_rl.algorithms.loss.loss_functions.DistillationLossDataDict attribute) teacher_vocab_sizes (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) teacher_weight_bytes (dockyard_rl.algorithms.distillation_memory.MemoryReport attribute) teacher_weights (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) teacher_xtoken_loss (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) TeacherConfig (class in dockyard_rl.algorithms.distillation) (class in dockyard_rl.models.policy.teacher_worker_group) TeacherPanelConfig (class in dockyard_rl.algorithms.opd) TeacherResourceConfig (class in dockyard_rl.algorithms.opd) TeacherScorer (class in dockyard_rl.models.policy.teacher_scorer) TeacherWorkerGroup (class in dockyard_rl.models.policy.teacher_worker_group) temperature (dockyard_rl.algorithms.logits_sampling_utils.TrainingSamplingParams attribute) (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) (dockyard_rl.models.generation.interfaces.GenerationConfig attribute) Tensor (in module dockyard_rl.data.llm_message_utils) tensor_parallel_size (dockyard_rl.algorithms.opd.TeacherResourceConfig attribute) (dockyard_rl.cluster.placement.ParallelismSpec attribute) (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) (dockyard_rl.models.jax.config.JaxPolicyConfig attribute) (dockyard_rl.models.policy.teacher_worker_group.TeacherConfig attribute) tensorboard (dockyard_rl.utils.logger.LoggerConfig attribute) tensorboard_enabled (dockyard_rl.utils.logger.LoggerConfig attribute) TensorboardConfig (class in dockyard_rl.utils.logger) TensorboardLogger (class in dockyard_rl.utils.logger) terminal (dockyard_rl.experience.cua.backends.interface.Observation attribute) terminal_bench_data_processor() (in module dockyard_rl.data.processors) TerminalBenchDataset (class in dockyard_rl.data.datasets.terminal_bench) TerminalBenchEnvConfig (class in dockyard_rl.environments.terminal_bench_environment) TerminalBenchEnvironment (class in dockyard_rl.environments.terminal_bench_environment) TerminalBenchReward (class in dockyard_rl.rewards.terminal_bench) terminateds (dockyard_rl.environments.interfaces.EnvironmentReturn attribute) test_command (dockyard_rl.sandbox.task_executor_client.TaskSpec attribute) test_patch (dockyard_rl.sandbox.task_executor_client.TaskSpec attribute) test_patch_applied (dockyard_rl.sandbox.task_executor_client.TaskResult attribute) TestRunnerReward (class in dockyard_rl.rewards.test_runner) tests_failed (dockyard_rl.sandbox.task_executor_client.TaskResult attribute) tests_passed (dockyard_rl.sandbox.task_executor_client.TaskResult attribute) text_column (dockyard_rl.environments.retriever.RAGEnvConfig attribute) TEXT_EXTENSIONS (in module dockyard_rl.environments._gdpval_extract) text_preprocessor() (in module dockyard_rl.data.processors) TheEyeClient (class in dockyard_rl.experience.cua.vision) THINKING_CLOSING_ONLY (in module dockyard_rl.tool_protocol.protocol) THINKING_PAIRED (in module dockyard_rl.tool_protocol.protocol) thinking_style (dockyard_rl.tool_protocol.protocol.StructuredToolUseConfig attribute) THINKING_STYLE_CLOSING_ONLY (in module dockyard_rl.rewards.invalid_action) THINKING_STYLE_PAIRED (in module dockyard_rl.rewards.invalid_action) thinking_tags (dockyard_rl.rewards.invalid_action.InvalidActionPenaltyConfig attribute) tie_word_embeddings (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) (dockyard_rl.models.jax.models.qwen3.Qwen3Config attribute) (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) (dockyard_rl.models.jax.moe.qwen3_moe.Qwen3MoeConfig attribute) TIGHT (dockyard_rl.algorithms.distillation_memory.ColocationSchedule attribute) time() (dockyard_rl.models.policy.interfaces.Timer method) (dockyard_rl.utils.timer.Timer method) timed_out (dockyard_rl.sandbox.task_executor_client.ExecResult attribute) (dockyard_rl.sandbox.task_executor_client.FinishResult attribute) timeout (dockyard_rl.sandbox.task_executor_client.TaskSpec attribute) TimeoutChecker (class in dockyard_rl.utils.timer) Timer (class in dockyard_rl.models.policy.interfaces) (class in dockyard_rl.utils.timer) to() (dockyard_rl.data.multimodal_utils.PackedTensor method) (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method)

to_dict() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) to_local_if_dtensor() (in module dockyard_rl.models.dtensor.parallelize) to_log_dict() (dockyard_rl.data.packing.metrics.PackingMetrics method) (dockyard_rl.data.packing.packing.PackingMetrics method) to_nested_by_length() (in module dockyard_rl.data_plane.codec) to_numpy() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) to_torch() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) token_indices_experts_sorted_N (dockyard_rl.models.dtensor.moe.dispatch.LocalDispatchMetadata attribute) (dockyard_rl.models.jax.moe.dispatch.LocalDispatchMetadata attribute) TOKEN_LEVEL (dockyard_rl.algorithms.loss.interfaces.LossType attribute) token_level_loss (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) token_mask (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossDataDict attribute) (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossDataDict attribute) (dockyard_rl.algorithms.loss.loss_functions.DistillationLossDataDict attribute) (dockyard_rl.algorithms.loss.loss_functions.DPOLossDataDict attribute) (dockyard_rl.algorithms.loss.loss_functions.DraftCrossEntropyLossDataDict attribute) (dockyard_rl.algorithms.loss.loss_functions.PreferenceLossDataDict attribute) TokenAligner (class in dockyard_rl.algorithms.x_token.token_aligner) TokenChoiceTopKRouter (class in dockyard_rl.models.dtensor.moe.router) (class in dockyard_rl.models.jax.moe.router) TokenizerType (in module dockyard_rl.data.collate_fn) (in module dockyard_rl.data.datasets.processed_dataset) (in module dockyard_rl.data.datasets.utils) (in module dockyard_rl.data.interfaces) (in module dockyard_rl.data.llm_message_utils) (in module dockyard_rl.data.processors) (in module dockyard_rl.experience.rollouts) tool_calls (dockyard_rl.tool_protocol.hermes.HermesExtraction attribute) tool_parser_plugin (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) ToolRegistry (class in dockyard_rl.tool_protocol.registry) tools (dockyard_rl.tool_protocol.registry.ToolRegistry attribute) tools_called (dockyard_rl.tool_protocol.hermes.HermesExtraction attribute) ToolUseAssessment (class in dockyard_rl.tool_protocol.registry) top_k (dockyard_rl.algorithms.logits_sampling_utils.TrainingSamplingParams attribute) (dockyard_rl.environments.retriever.RAGEnvConfig attribute) (dockyard_rl.models.generation.interfaces.GenerationConfig attribute) top_k_experts (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) TOP_K_THRESHOLD (in module dockyard_rl.models.generation.sglang.sglang_generation) TOP_K_TOP_P_CHUNK_SIZE (in module dockyard_rl.algorithms.logits_sampling_utils) top_p (dockyard_rl.algorithms.logits_sampling_utils.TrainingSamplingParams attribute) (dockyard_rl.models.generation.interfaces.GenerationConfig attribute) TOP_P_THRESHOLD (in module dockyard_rl.models.generation.sglang.sglang_generation) topk_indices (dockyard_rl.models.policy.interfaces.TopkLogitsOutputSpec attribute) topk_logits (dockyard_rl.models.policy.interfaces.TopkLogitsOutputSpec attribute) topk_scores_experts_sorted_N (dockyard_rl.models.dtensor.moe.dispatch.LocalDispatchMetadata attribute) (dockyard_rl.models.jax.moe.dispatch.LocalDispatchMetadata attribute) TopkLogitsOutputSpec (class in dockyard_rl.models.policy.interfaces) TOPO_RANK_KEY (in module dockyard_rl.distributed.virtual_cluster) TOPO_RANK_UNKNOWN (in module dockyard_rl.distributed.virtual_cluster) torch_compile_max_bs (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) torch_to_jnp() (in module dockyard_rl.models.jax.convert) torchao_config (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) total_bytes (dockyard_rl.data_plane.observability.DataPlaneStats attribute) total_gpus (dockyard_rl.cluster.placement.ParallelismSpec property) total_keys (dockyard_rl.data_plane.observability.DataPlaneStats attribute) total_ops (dockyard_rl.data_plane.observability.DataPlaneStats attribute) total_packs (dockyard_rl.data.packing.metrics.PackingMetrics attribute) (dockyard_rl.data.packing.packing.PackingMetrics attribute) total_padding (dockyard_rl.data.packing.metrics.PackingMetrics attribute) (dockyard_rl.data.packing.packing.PackingMetrics attribute) total_sequences (dockyard_rl.data.packing.metrics.PackingMetrics attribute) (dockyard_rl.data.packing.packing.PackingMetrics attribute) total_steps (dockyard_rl.algorithms.distillation.DistillationSaveState attribute) (dockyard_rl.algorithms.dpo.DPOSaveState attribute) (dockyard_rl.algorithms.grpo.GRPOSaveState attribute) (dockyard_rl.algorithms.rm.RMSaveState attribute) (dockyard_rl.algorithms.sft.SFTSaveState attribute) total_tokens (dockyard_rl.data.packing.metrics.PackingMetrics attribute) (dockyard_rl.data.packing.packing.PackingMetrics attribute) total_valid_tokens (dockyard_rl.algorithms.distillation.DistillationSaveState attribute) (dockyard_rl.algorithms.dpo.DPOSaveState attribute) (dockyard_rl.algorithms.grpo.GRPOSaveState attribute) (dockyard_rl.algorithms.rm.RMSaveState attribute) (dockyard_rl.algorithms.sft.SFTSaveState attribute) TP tp (dockyard_rl.models.dtensor.moe.mesh.MoEParallelDims attribute) (dockyard_rl.models.jax.sharding.MoEMeshDims attribute) TQDataPlaneClient (class in dockyard_rl.data_plane.adapters.transfer_queue) TQPolicy (class in dockyard_rl.models.policy.tq_policy) TQWorkerMixin (class in dockyard_rl.data_plane.worker_mixin) tracking_uri (dockyard_rl.utils.logger.MLflowConfig attribute) train (dockyard_rl.data.DataConfig attribute) train() (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.interfaces.PolicyInterface method) (dockyard_rl.models.policy.lm_policy.Policy method) (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) train_context() (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl static method) train_from_meta() (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl method) (dockyard_rl.models.policy.tq_policy.TQPolicy method) train_presharded() (dockyard_rl.data_plane.worker_mixin.TQWorkerMixin method) train_step() (in module dockyard_rl.models.jax.train_step) trainer_bundle_ct_per_node() (in module dockyard_rl.cluster.placement) trainer_sync_ranks() (in module dockyard_rl.cluster.placement) TrainFn (in module dockyard_rl.algorithms.online_dpo) training_loss_fn() (in module dockyard_rl.models.jax.train_step) training_phase_bytes (dockyard_rl.algorithms.distillation_memory.MemoryReport property) TrainingSamplingParams (class in dockyard_rl.algorithms.logits_sampling_utils) trajectory (dockyard_rl.algorithms.online_dpo.Candidate attribute) translate_parallel_style() (in module dockyard_rl.models.dtensor.parallelize) triton_attention_num_kv_splits (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) triton_attention_reduce_in_fp32 (dockyard_rl.models.generation.sglang.config.SglangSpecificArgs attribute) truncated (dockyard_rl.models.generation.interfaces.GenerationOutputSpec attribute) truncated_importance_sampling_ratio (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) truncated_importance_sampling_ratio_min (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) truncated_importance_sampling_type (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) Tulu3PreferenceDataset (class in dockyard_rl.data.datasets.preference_datasets.tulu3) Tulu3SftMixtureDataset (class in dockyard_rl.data.datasets.response_datasets.tulu3) type (dockyard_rl.rewards.invalid_action.Violation attribute)

U

uncommon_topk (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute) undesirable_weight (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.loss.loss_functions.DPOLossConfig attribute) unpack_tensor() (in module dockyard_rl.models.huggingface.common) unpadded_sequence_lengths (dockyard_rl.models.generation.interfaces.GenerationOutputSpec attribute) unshard_fsdp2_model() (in module dockyard_rl.models.policy.workers.dtensor_policy_worker) unwrap_wire_stripped_payload() (in module dockyard_rl.data_plane.codec) update() (dockyard_rl.algorithms.online_dpo.JudgeOveroptMonitor method) (dockyard_rl.data.packing.metrics.PackingMetrics method) (dockyard_rl.data.packing.packing.PackingMetrics method) (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) update_expert_biases() (in module dockyard_rl.models.dtensor.moe.load_balance) (in module dockyard_rl.models.jax.moe.load_balance) update_single_dataset_config() (in module dockyard_rl.data.datasets.utils) update_weights_from_collective() (dockyard_rl.models.generation.interfaces.GenerationInterface method) (dockyard_rl.models.generation.sglang.sglang_generation.SGLangGeneration method) (dockyard_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (dockyard_rl.models.generation.vllm.vllm_generation.VllmGeneration method) (dockyard_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) (dockyard_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method)

update_weights_via_ipc_zmq() (dockyard_rl.models.generation.interfaces.GenerationInterface method) (dockyard_rl.models.generation.sglang.sglang_generation.SGLangGeneration method) (dockyard_rl.models.generation.vllm.vllm_backend.VllmInternalWorkerExtension method) (dockyard_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) use_activation_pow2_scale (dockyard_rl.models.generation.vllm.quantization.fp8.FP8Config attribute) use_bidirectional_attention (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) use_cispo (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) use_deep_gemm (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) use_double_wide_mlp (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) use_dynamic_sampling (dockyard_rl.algorithms.grpo.GRPOConfig attribute) use_environment_flags (dockyard_rl.rewards.invalid_action.InvalidActionPenaltyConfig attribute) use_fp8_weights (dockyard_rl.models.generation.vllm.quantization.fp8.FP8Config attribute) use_importance_sampling_correction (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) use_kl_in_reward (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) use_leave_one_out_baseline (dockyard_rl.algorithms.grpo.AdvEstimatorConfig attribute) (dockyard_rl.algorithms.grpo.GRPOConfig attribute) use_multiple_dataloader (dockyard_rl.data.DataConfig attribute) use_on_policy_kl_approximation (dockyard_rl.algorithms.loss.loss_functions.ClippedPGLossConfig attribute) use_reference_model() (dockyard_rl.models.policy.workers.dtensor_policy_worker.DTensorPolicyWorkerImpl method) use_weight_pow2_scale (dockyard_rl.models.generation.vllm.quantization.fp8.FP8Config attribute)

V

V (in module dockyard_rl.experience.cua.vendor.desktop_env_eval.metrics.utils) val_at_end (dockyard_rl.algorithms.distillation.DistillationConfig attribute) (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.grpo.GRPOConfig attribute) (dockyard_rl.algorithms.rm.RMConfig attribute) (dockyard_rl.algorithms.sft.SFTConfig attribute) val_at_start (dockyard_rl.algorithms.distillation.DistillationConfig attribute) (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.grpo.GRPOConfig attribute) (dockyard_rl.algorithms.rm.RMConfig attribute) (dockyard_rl.algorithms.sft.SFTConfig attribute) val_batch_size (dockyard_rl.algorithms.grpo.GRPOConfig attribute) val_batches (dockyard_rl.algorithms.distillation.DistillationConfig attribute) (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.rm.RMConfig attribute) (dockyard_rl.algorithms.sft.SFTConfig attribute) val_dataset (dockyard_rl.data.datasets.raw_dataset.RawDataset attribute) val_global_batch_size (dockyard_rl.algorithms.distillation.DistillationConfig attribute) (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.rm.RMConfig attribute) (dockyard_rl.algorithms.sft.SFTConfig attribute) val_loss (dockyard_rl.algorithms.distillation.DistillationSaveState attribute) val_micro_batch_size (dockyard_rl.algorithms.distillation.DistillationConfig attribute) (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.rm.RMConfig attribute) (dockyard_rl.algorithms.sft.SFTConfig attribute) val_period (dockyard_rl.algorithms.distillation.DistillationConfig attribute) (dockyard_rl.algorithms.dpo.DPOConfig attribute) (dockyard_rl.algorithms.grpo.GRPOConfig attribute) (dockyard_rl.algorithms.rm.RMConfig attribute) (dockyard_rl.algorithms.sft.SFTConfig attribute) val_reward (dockyard_rl.algorithms.grpo.GRPOSaveState attribute) valid_calls (dockyard_rl.tool_protocol.registry.ToolUseAssessment attribute) valid_chunk_mask() (in module dockyard_rl.algorithms.x_token.loss_utils) validate() (in module dockyard_rl.algorithms.distillation) (in module dockyard_rl.algorithms.dpo) (in module dockyard_rl.algorithms.grpo) (in module dockyard_rl.algorithms.rm) (in module dockyard_rl.algorithms.sft) validate_against_schema() (in module dockyard_rl.tool_protocol.schema) validate_arguments() (in module dockyard_rl.tool_protocol.schema) validate_devices() (dockyard_rl.models.jax.sharding.MoEMeshDims method) validate_effort_reward_config() (in module dockyard_rl.algorithms.effort_reward) validate_for_inference() (in module dockyard_rl.cluster.placement) validate_for_trainer() (in module dockyard_rl.cluster.placement) validate_gguf() (in module dockyard_rl.utils.gguf_export) validate_graded_allocation_config() (in module dockyard_rl.algorithms.rollout_allocation) validate_moe_parallel_config() (in module dockyard_rl.models.dtensor.moe.config) validate_one_dataset() (in module dockyard_rl.algorithms.dpo) (in module dockyard_rl.algorithms.rm) validate_packing_algorithm() (in module dockyard_rl.data.packing.algorithms) validate_precision() (in module dockyard_rl.utils.device_capability) validate_router_replay_config() (in module dockyard_rl.models.dtensor.moe.router_replay) validate_router_replay_generation_compat() (in module dockyard_rl.models.dtensor.moe.router_replay) validate_routing_diversity_config() (in module dockyard_rl.algorithms.routing_diversity) validate_task_config() (in module dockyard_rl.experience.cua.tasks.validate) validate_teacher_panel_config() (in module dockyard_rl.algorithms.teacher_panel) validate_tree() (in module dockyard_rl.experience.cua.tasks.validate) validation (dockyard_rl.data.DataConfig attribute)

values() (dockyard_rl.distributed.batched_data_dict.BatchedDataDict method) variables_after_stage (dockyard_rl.utils.memory_tracker.MemoryTrackerDataPoint attribute) variables_before_stage (dockyard_rl.utils.memory_tracker.MemoryTrackerDataPoint attribute) variance_exponent (dockyard_rl.algorithms.rollout_allocation.GradedAllocationConfig attribute) verifier_type (dockyard_rl.environments.math_environment.MathEnvConfig attribute) verify() (dockyard_rl.environments.code_jaccard_environment.CodeJaccardVerifyWorker method) (dockyard_rl.environments.math_environment.EnglishMultichoiceVerifyWorker method) (dockyard_rl.environments.math_environment.HFMultiRewardVerifyWorker method) (dockyard_rl.environments.math_environment.HFVerifyWorker method) (dockyard_rl.environments.math_environment.MultilingualMultichoiceVerifyWorker method) (dockyard_rl.environments.vlm_environment.VLMVerifyWorker method) (in module dockyard_rl.environments.dapo_math_verifier) verify_right_padding() (in module dockyard_rl.models.generation.interfaces) Violation (class in dockyard_rl.rewards.invalid_action) VIOLATION_ENV_INVALID_ACTION (in module dockyard_rl.rewards.invalid_action) VIOLATION_ENV_MALFORMED_THINKING (in module dockyard_rl.rewards.invalid_action) VIOLATION_MALFORMED_THINKING (in module dockyard_rl.rewards.invalid_action) VIOLATION_MALFORMED_TOOL_CALL (in module dockyard_rl.rewards.invalid_action) violation_penalty() (in module dockyard_rl.rewards.invalid_action) violation_rate_metrics() (in module dockyard_rl.rewards.invalid_action) VIOLATION_SCHEMA_ARGS (in module dockyard_rl.rewards.invalid_action) VIOLATION_SCHEMA_UNKNOWN_TOOL (in module dockyard_rl.rewards.invalid_action) VIOLATION_UNEXECUTED_PATTERN (in module dockyard_rl.rewards.invalid_action) violations (dockyard_rl.rewards.invalid_action.InvalidActionVerdict attribute) violations_at_locus() (in module dockyard_rl.rewards.invalid_action) VISIBLE_DEVICE_ENV_VARS (in module dockyard_rl.utils.device_query) visible_devices_entries() (in module dockyard_rl.utils.device_query) visible_devices_env() (in module dockyard_rl.utils.device_query) visible_devices_value() (in module dockyard_rl.utils.device_query) VISUAL_BYTE_MAP (in module dockyard_rl.algorithms.x_token.token_aligner) VLLM_BACKEND (in module dockyard_rl.models.generation.constants) vllm_cfg (dockyard_rl.models.generation.vllm.config.VllmConfig attribute) vllm_kwargs (dockyard_rl.models.generation.vllm.config.VllmConfig attribute) VLLM_LOAD_FORMAT_AUTO (dockyard_rl.models.huggingface.common.ModelFlag attribute) vllm_metrics_logger_interval (dockyard_rl.models.generation.vllm.config.VllmSpecificArgs attribute) vllm_patches (dockyard_rl.models.generation.vllm.quantization.fp8.FP8State attribute) VllmAsyncGenerationWorkerImpl (class in dockyard_rl.models.generation.vllm.vllm_worker_async) VllmConfig (class in dockyard_rl.models.generation.vllm.config) VllmGeneration (class in dockyard_rl.models.generation.vllm.vllm_generation) VllmGenerationWorker (class in dockyard_rl.models.generation.vllm.vllm_worker) VllmGenerationWorkerImpl (class in dockyard_rl.models.generation.vllm.vllm_worker) VllmInternalWorkerExtension (class in dockyard_rl.models.generation.vllm.vllm_backend) VllmSpecificArgs (class in dockyard_rl.models.generation.vllm.config) vlm_hf_data_processor() (in module dockyard_rl.data.processors) VLMEnvConfig (class in dockyard_rl.environments.vlm_environment) VLMEnvironment (class in dockyard_rl.environments.vlm_environment) VLMEnvironmentMetadata (class in dockyard_rl.environments.vlm_environment) VLMVerifyWorker (class in dockyard_rl.environments.vlm_environment) vm_ip (dockyard_rl.experience.cua.vendor.desktop_env_eval.grading_env.GradingEnv attribute) vocab_parallel_argmax() (in module dockyard_rl.distributed.model_utils) vocab_parallel_full_log_softmax() (in module dockyard_rl.distributed.model_utils) vocab_parallel_gather_columns() (in module dockyard_rl.distributed.model_utils) vocab_parallel_group (dockyard_rl.algorithms.loss.loss_functions.DraftCrossEntropyLossConfig attribute) vocab_parallel_log_softmax() (in module dockyard_rl.distributed.model_utils) vocab_size (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) (dockyard_rl.models.jax.models.qwen3.Qwen3Config attribute) (dockyard_rl.models.jax.models.qwen3_next.Qwen3NextConfig attribute) (dockyard_rl.models.jax.moe.qwen3_moe.Qwen3MoeConfig attribute) vocab_size_per_layer_input (dockyard_rl.models.jax.models.gemma4.Gemma4TextConfig attribute) vocab_topk (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute)

W

wake_up() (dockyard_rl.models.generation.sglang.sglang_worker.SGLangGenerationWorker method) (dockyard_rl.models.generation.vllm.vllm_worker.VllmGenerationWorkerImpl method) wake_up_async() (dockyard_rl.models.generation.vllm.vllm_worker_async.VllmAsyncGenerationWorkerImpl method) walk_files() (in module dockyard_rl.environments._gdpval_extract) wall_ms (dockyard_rl.data_plane.observability.DataPlaneEvent attribute) wandb (dockyard_rl.utils.logger.LoggerConfig attribute) wandb_enabled (dockyard_rl.utils.logger.LoggerConfig attribute) WandbConfig (class in dockyard_rl.utils.logger) WandbLogger (class in dockyard_rl.utils.logger) weight (dockyard_rl.rewards.gdpval_rubric.RubricItem attribute) WeightSynchronizer (class in dockyard_rl.weight_sync.interfaces) WindowsBackend (class in dockyard_rl.experience.cua.backends.windows) wire_ep() (dockyard_rl.models.jax.moe.alltoall.AllToAllTokenDispatcher method) wire_ep_mesh() (dockyard_rl.models.dtensor.moe.dispatch.AllToAllTokenDispatcher method)

WORKER_CLASS_DICT (dockyard_rl.environments.math_environment.BaseMathEnvironment attribute) (dockyard_rl.environments.math_environment.MathEnvironment attribute) (dockyard_rl.environments.math_environment.MathMultiRewardEnvironment attribute) worker_group (dockyard_rl.models.jax.policy_worker.JaxPolicyWorkerImpl attribute) (dockyard_rl.models.policy.interfaces.PolicyInterface attribute) worker_metadata (dockyard_rl.distributed.worker_groups.RayWorkerGroup property) workers (dockyard_rl.distributed.worker_groups.RayWorkerGroup property) world_size (dockyard_rl.models.dtensor.moe.mesh.MoEParallelDims attribute) (dockyard_rl.models.jax.sharding.MoEMeshDims property) world_size() (dockyard_rl.distributed.virtual_cluster.RayVirtualCluster method) wrap_loss_fn_with_input_preparation() (in module dockyard_rl.algorithms.loss.wrapper) wrap_with_nvtx_name() (in module dockyard_rl.models.generation.vllm.vllm_worker) (in module dockyard_rl.utils.nsys) write_columns() (in module dockyard_rl.data_plane.column_io) write_to_dataplane() (dockyard_rl.models.policy.tq_policy.TQPolicy method)

X

XLA	xtoken xtoken_loss (dockyard_rl.algorithms.loss.loss_functions.CrossTokenizerDistillationLossConfig attribute)

Z

zero_outside_topk (dockyard_rl.algorithms.distillation.DistillationConfig attribute) (dockyard_rl.algorithms.loss.loss_functions.DistillationLossConfig attribute)

ZMQ