MCPcopy Create free account

hub / github.com/Gen-Verse/OpenClaw-RL / functions

Functions10,561 in github.com/Gen-Verse/OpenClaw-RL

↓ 1 callersFunction_read_megatron_common_args
Read the `args` Namespace from a Megatron torch_dist checkpoint's common.pt. Returns None if the path is not a Megatron checkpoint or common.pt i
slime/slime/backends/megatron_utils/actor.py:98
↓ 1 callersMethod_reap_idle_locked
(self)
gui-rl/env_pool_server.py:152
↓ 1 callersMethod_record_conversation_turn
Record conversation turns for experience extraction later.
openclaw-rl/oel/openclaw_oel_api_server.py:1038
↓ 1 callersMethod_recv_task
Listens for completed inference requests from the coordinator. This coroutine runs in an infinite loop, continuously polling the soc
Megatron-LM/megatron/core/inference/inference_client.py:119
↓ 1 callersFunction_redo_attention_load_balancing
Redo the context parallel attention load balancing For example, for cp_size=3, converts 123456 to 162534 for efficient processing by atte
Megatron-LM/megatron/core/ssm/mamba_context_parallel.py:376
↓ 1 callersMethod_register_default_tools
Register default tools in the registry
toolcall-rl/tool_sandbox.py:310
↓ 1 callersMethod_register_fsdp_hooks
Register necessary hooks for Fully Sharded Data Parallel (FSDP) execution on the model. This function sets up various hooks required for FSDP
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/megatron_fsdp.py:457
↓ 1 callersMethod_register_hooks_on_fused_impl
Attempt to emulate submodule callback hooks. This is not always possible because Transformer Engine's op fuser does not expos
Megatron-LM/megatron/core/extensions/transformer_engine.py:1839
↓ 1 callersMethod_register_load_state_dict_hooks
(self)
Megatron-LM/megatron/core/optimizer/cpu_offloading/hybrid_optimizer.py:383
↓ 1 callersMethod_register_param_copy_back_gpu_hook
(self)
Megatron-LM/megatron/core/optimizer/cpu_offloading/hybrid_optimizer.py:117
↓ 1 callersMethod_relative_position_bucket
Adapted from HuggingFace T5 Model: https://github.com/huggingface/transformers/blob/329f5dbf97a5cb2473914c88c05aa3dcb242e19a/
Megatron-LM/megatron/core/models/common/embeddings/relative_pos_embedding.py:46
↓ 1 callersMethod_release_state
Clear the state of the node
Megatron-LM/megatron/core/pipeline_parallel/utils.py:287
↓ 1 callersFunction_remove_msc_prefix
Remove the MSC prefix from a path Args: path (str): The path Returns: str: The path without the MSC prefix
Megatron-LM/megatron/core/datasets/object_storage_utils.py:94
↓ 1 callersMethod_remove_node_and_descendants
Remove a node and all its descendants from the trie. Args: node: The node to remove along with all its descendants
slime/slime/router/middleware_hub/radix_tree.py:400
↓ 1 callersMethod_remove_node_from_parent
Remove a node from its parent's children list.
slime/slime/router/middleware_hub/radix_tree.py:431
↓ 1 callersMethod_remove_redundant_data
( fully_parallel, sharded_part, shard_to_saving_rank, parallelization_group )
Megatron-LM/megatron/core/dist_checkpointing/tensor_aware_state_dict.py:86
↓ 1 callersFunction_remove_right_units
(string)
slime/slime/rollout/rm_hub/math_utils.py:75
↓ 1 callersFunction_remove_s3_prefix
Remove the S3 prefix from a path Args: path (str): The path Returns: str: The path without the S3 prefix
Megatron-LM/megatron/core/datasets/object_storage_utils.py:70
↓ 1 callersFunction_replace_sharded_keys_with_state_dict_keys
Inverse of _replace_state_dict_keys_with_sharded_keys.
Megatron-LM/megatron/core/dist_checkpointing/strategies/torch.py:431
↓ 1 callersMethod_replay_session
Replay mode: extract experience from the full session, then run teacher on every turn. Flow: 1. Extract one experience from the ent
openclaw-rl/oel/openclaw_oel_api_server.py:938
↓ 1 callersMethod_report_quantize_tensor_info
(self)
Megatron-LM/megatron/core/post_training/modelopt/layers.py:233
↓ 1 callersMethod_reset
(self, input_message: Any)
terminal-rl/agent_runner.py:37
↓ 1 callersFunction_resolve_callable_from_python_import_path
Resolve a Python import path like 'pkg.mod.func' to a callable. Raises ValueError with clear message on failure.
Megatron-LM/megatron/core/fp8_utils.py:121
↓ 1 callersFunction_resolve_eval_datasets
Build evaluation dataset configurations from either --eval-config or --eval-prompt-data.
slime/slime/utils/arguments.py:1758
↓ 1 callersFunction_resolve_pull_image
(task: dict[str, Any])
terminal-rl/remote/docker_compose_utils.py:131
↓ 1 callersFunction_resolve_scheduler_role_arn
(logger)
gui-rl/desktop_env/providers/aws/scheduler_utils.py:9
↓ 1 callersFunction_response_logprobs_hf
( model, input_ids_1d: torch.Tensor, response_len: int, device: torch.device, temperature:
openclaw-combine/prm_teacher_postprocess.py:91
↓ 1 callersFunction_restore_dict_types
Recursively update `x` keys, based on `keys_template`.
Megatron-LM/megatron/core/dist_checkpointing/strategies/torch.py:446
↓ 1 callersMethod_restore_model
Returns tokenizer library object.
Megatron-LM/megatron/core/tokenizers/text/text_tokenizer.py:52
↓ 1 callersMethod_revert_to_snapshot
(self)
gui-rl/desktop_env/desktop_env.py:222
↓ 1 callersFunction_roll_tensor_packed_seq
Roll tensor with packed sequence support. This function handles rolling for packed sequences by respecting sequence boundaries
Megatron-LM/megatron/core/transformer/multi_token_prediction.py:228
↓ 1 callersFunction_rotate_half
Change sign so the last dimension becomes [-odd, +even] Args: x (Tensor): Input tensor Returns: Tensor: Tensor rotated half
Megatron-LM/megatron/core/models/common/embeddings/rope_utils.py:73
↓ 1 callersFunction_run_agent_remote
Multi-turn agent loop. Produces per-message token_ids that guarantee byte-perfect consistency between rollout-time context and training-time token
swe-rl/generate_with_swe_trajectory.py:570
↓ 1 callersMethod_run_health_checks
(self)
slime/slime/utils/health_monitor.py:138
↓ 1 callersFunction_run_inference_step
(url: str, tokens: list[int], sampling_params: dict, image_data, tokenizer)
slime/examples/geo3k_vlm_multi_turn/rollout.py:192
↓ 1 callersMethod_run_model_turn
( self, context_messages: List[Dict[str, Any]] )
terminal-rl/agent_runner.py:51
↓ 1 callersFunction_run_one
(row, row_index: int, repeat_index: int)
slime/slime/utils/debug_utils/send_to_sglang.py:38
↓ 1 callersMethod_run_split_on_punc
Splits punctuation on a piece of text.
Megatron-LM/megatron/training/tokenizer/bert_tokenization.py:263
↓ 1 callersMethod_run_strip_accents
Strips accents from a piece of text.
Megatron-LM/megatron/training/tokenizer/bert_tokenization.py:252
↓ 1 callersFunction_s3_download_file
Download the object at the given S3 path to the given local file system path Args: client (S3Client): The S3 client s3_path (str
Megatron-LM/megatron/core/datasets/object_storage_utils.py:119
↓ 1 callersMethod_safe_create_task
(self, coro)
openclaw-rl/openclaw_api_server.py:664
↓ 1 callersMethod_safe_create_task
(self, coro)
openclaw-opd/openclaw_opd_api_server.py:938
↓ 1 callersFunction_sample_task_info
(sample: Sample, evaluation: bool = False)
gui-rl/generate_with_gui.py:91
↓ 1 callersFunction_sanitize_filename
(value: str)
swe-rl/generate_with_swe_trajectory.py:105
↓ 1 callersMethod_save_overview_data_yaml
Save a yaml report of the instances and their exit statuses.
swe-rl/mini-swe-agent/src/minisweagent/run/extra/utils/batch_progress.py:175
↓ 1 callersFunction_save_rollout_artifacts
Best-effort trajectory dump to ``SWE_SAVE_TRAJ_DIR`` (optional).
swe-rl/generate_with_swe_trajectory.py:119
↓ 1 callersMethod_save_session_trajectory
Save a session's trajectory data to disk for later consolidation.
openclaw-rl/oel/openclaw_oel_api_server.py:758
↓ 1 callersMethod_save_state
Internal method that saves the state that needs to be restored when rewound. Any state that may change during the execution of a step before
Megatron-LM/megatron/core/rerun_state_machine.py:887
↓ 1 callersFunction_save_to_existing_array
(sharded_tensor: ShardedTensor, arr: Optional[zarr.Array])
Megatron-LM/megatron/core/dist_checkpointing/strategies/zarr.py:147
↓ 1 callersMethod_score_answer
Use the same logic as the single-turn math reward model. We accept either boxed or raw numeric strings by retrying with a boxed wrapp
slime/examples/geo3k_vlm_multi_turn/env_geo3k.py:88
↓ 1 callersMethod_search
(self, query: str, num: int, return_score: bool)
slime/examples/search-r1/local_dense_retriever/retrieval_server.py:132
↓ 1 callersMethod_search
(self, query: str, num: int = None, return_score: bool = False)
slime/examples/search-r1/local_dense_retriever/retrieval_server.py:159
↓ 1 callersFunction_select_best_hint
(votes: list[dict[str, Any]])
openclaw-opd/openclaw_opd_api_server.py:130
↓ 1 callersFunction_select_best_hint
(votes: list[dict[str, Any]])
openclaw-fireworks/fireworks_server.py:217
↓ 1 callersMethod_select_layers_for_pipeline_parallel
(self, layer_type_list)
Megatron-LM/megatron/core/ssm/mamba_block.py:168
↓ 1 callersMethod_select_teacher_cand_per_sample
Pick one candidate per token / per step / per sample by overlap. For each sample i: * student_topk_idx[i]: [R_i, K_q] * t
slime/slime/backends/megatron_utils/actor.py:838
↓ 1 callersMethod_send_hf_params
(self, hf_named_tensors)
slime/slime/backends/megatron_utils/update_weight/update_weight_from_tensor.py:145
↓ 1 callersFunction_send_to_colocated_engine
( hf_named_tensors: list[tuple[str, torch.Tensor]], *, ipc_engine, ipc_gather_src, ipc_gat
slime/slime/backends/megatron_utils/update_weight/update_weight_from_tensor.py:171
↓ 1 callersMethod_sequence_pointers
Build the sequence pointers per the sequence lengths and dtype size Args: sequence_lengths (List[int]): The length of each sequen
Megatron-LM/megatron/core/datasets/indexed_dataset.py:212
↓ 1 callersFunction_set_adlr_autoresume
Initialize ADLR autoresume.
Megatron-LM/megatron/training/global_vars.py:246
↓ 1 callersMethod_set_attention_backend
Set attention backend Transformer engine works based on optout. By default all three attention backend flags are set to 1. So if the user cho
Megatron-LM/megatron/core/models/common/language_module/language_module.py:93
↓ 1 callersFunction_set_cache_control
(entry: dict)
swe-rl/mini-swe-agent/src/minisweagent/models/utils/cache_control.py:15
↓ 1 callersMethod_set_deep_tensor
(cls, obj, path, src)
Megatron-LM/tools/checkpoint/schema_base.py:43
↓ 1 callersFunction_set_energy_monitor
Initialize energy monitor.
Megatron-LM/megatron/training/global_vars.py:270
↓ 1 callersMethod_set_evaluator_info
Set evaluator information from task config
gui-rl/desktop_env/desktop_env.py:334
↓ 1 callersFunction_set_global_memory_buffer
Initialize global buffer.
Megatron-LM/megatron/core/parallel_state.py:1978
↓ 1 callersFunction_set_one_logger
(args)
Megatron-LM/megatron/training/global_vars.py:222
↓ 1 callersFunction_set_random_seed
Set random seed for reproducability.
slime/slime/backends/megatron_utils/initialize.py:14
↓ 1 callersFunction_set_rerun_state_machine
Internal function to set the singleton instance of the rerun machine.
Megatron-LM/megatron/core/rerun_state_machine.py:1305
↓ 1 callersFunction_set_signal_handler
(exit_signal)
Megatron-LM/megatron/training/global_vars.py:77
↓ 1 callersMethod_set_sub_optimizer_grads
(self)
Megatron-LM/megatron/core/optimizer/cpu_offloading/hybrid_optimizer.py:83
↓ 1 callersMethod_set_task_info
Set task info (proxy logic is handled in reset method)
gui-rl/desktop_env/desktop_env.py:324
↓ 1 callersFunction_set_tensorboard_writer
Set tensorboard writer.
Megatron-LM/megatron/training/global_vars.py:170
↓ 1 callersFunction_set_timers
Initialize timers.
Megatron-LM/megatron/training/global_vars.py:264
↓ 1 callersFunction_set_wandb_writer
(args)
Megatron-LM/megatron/training/global_vars.py:190
↓ 1 callersMethod_setup_attributes
( self, tiled_mma: cute.TiledMma, a_dtype: Type[cutlass.Numeric],
Megatron-LM/megatron/core/fusions/linear_cross_entropy/blackwell/fwd_mainloop.py:110
↓ 1 callersMethod_setup_attributes
( self, tiled_mma: cute.TiledMma, a_dtype: Type[cutlass.Numeric],
Megatron-LM/megatron/core/fusions/linear_cross_entropy/blackwell/bwd_partial_dlogits.py:97
↓ 1 callersMethod_setup_device_mesh
Setup device mesh for data parallelism.
slime/slime/backends/fsdp_utils/actor.py:216
↓ 1 callersFunction_setup_root_logger
()
swe-rl/mini-swe-agent/src/minisweagent/utils/log.py:7
↓ 1 callersMethod_setup_routes
Setup all the HTTP routes
slime/slime/router/router.py:66
↓ 1 callersFunction_shard_size
Returns size in bytes of a given sharded tensor.
Megatron-LM/megatron/core/dist_checkpointing/exchange_utils.py:63
↓ 1 callersMethod_should_call_local_cudagraph
Check if we should call the local cudagraph path.
Megatron-LM/megatron/core/transformer/transformer_layer.py:1006
↓ 1 callersMethod_should_call_local_cudagraph
Check if we should call the local cudagraph path.
Megatron-LM/megatron/core/transformer/transformer_block.py:547
↓ 1 callersMethod_should_call_local_cudagraph
Check if we should call the local cudagraph path.
Megatron-LM/megatron/core/transformer/module.py:286
↓ 1 callersMethod_should_call_te_cudagraph
Check if we should call the TE cudagraph path.
Megatron-LM/megatron/core/transformer/module.py:292
↓ 1 callersFunction_should_gather_qwen35_sequence_parallel_logits
()
slime/slime_plugins/megatron_bridge/qwen35_vl.py:50
↓ 1 callersFunction_should_log_qwen35_bridge_debug
()
slime/slime/backends/megatron_utils/checkpoint.py:14
↓ 1 callersFunction_should_stop_on_finish
(sample: Sample, finish_type: str)
slime/examples/geo3k_vlm_multi_turn/rollout.py:283
↓ 1 callersFunction_single_reward
(sample: Sample)
gui-rl/generate_with_gui.py:721
↓ 1 callersFunction_single_vote
(vote_id: int)
toolcall-rl/generate_with_retool.py:477
↓ 1 callersFunction_split_by_separation_hint
Splits buckets into those whose keys begin with the separation_hint and those whose keys do not Args: buckets (List[List[WriteItem]]
Megatron-LM/megatron/core/dist_checkpointing/strategies/filesystem_async.py:582
↓ 1 callersFunction_split_by_size_and_type
Splits write items according to item size into close to uniform bins. Same as torch.distributed.checkpoint.filesystem._split_by_size_and_typ
Megatron-LM/megatron/core/dist_checkpointing/strategies/filesystem_async.py:536
↓ 1 callersMethod_split_extra_state
(self, state)
Megatron-LM/megatron/core/extensions/kitchen.py:698
↓ 1 callersMethod_split_extra_state
(self, state)
Megatron-LM/megatron/core/extensions/transformer_engine.py:1458
↓ 1 callersFunction_split_linear_attn_in_proj_weight
(args, param)
slime/slime/backends/megatron_utils/megatron_to_hf/qwen3_5.py:140
↓ 1 callersFunction_split_qkv_bias
(args, param, head_dim, value_num_per_group)
slime/slime/backends/megatron_utils/megatron_to_hf/qwen3_5.py:127
↓ 1 callersFunction_split_qkv_weight
(args, param, head_dim, value_num_per_group)
slime/slime/backends/megatron_utils/megatron_to_hf/qwen3_5.py:112
↓ 1 callersMethod_split_state_dict
Split the state dict into sub-state dicts according to the chunks of each sub-optimizer in this chained optimizer. For example, assum
Megatron-LM/megatron/core/optimizer/optimizer.py:1143
↓ 1 callersMethod_split_train_data_by_dp
Split the train data by data parallel size.
slime/slime/ray/rollout.py:757
↓ 1 callersFunction_squared_relu_back
Gradient of Squared-ReLU. The derivative of ``(ReLU(x))^2`` w.r.t ``x`` is ``2 * ReLU(x)``.
Megatron-LM/megatron/core/fusions/fused_weighted_squared_relu.py:32
← previousnext →3,201–3,300 of 10,561, ranked by callers