MCPcopy Create free account

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

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

↓ 1 callersMethodcleanup
(self)
swe-rl/mini-swe-agent/src/minisweagent/environments/singularity.py:92
↓ 1 callersFunctioncleanup_memory
Force garbage collection to free memory
toolcall-rl/tool_sandbox.py:47
↓ 1 callersFunctioncleanup_old_non_persistent_checkpoint
(save_dir, leave_ckpt_num=1, do_async=False)
Megatron-LM/megatron/training/checkpointing.py:741
↓ 1 callersMethodclear_all
()
slime/slime/utils/routing_replay.py:47
↓ 1 callersMethodclear_all_forward
()
slime/slime/utils/routing_replay.py:52
↓ 1 callersMethodclear_forward
(self)
slime/slime/utils/routing_replay.py:43
↓ 1 callersMethodclear_memory
(self)
slime/slime/ray/train_actor.py:98
↓ 1 callersFunctionclear_output_dir
Clear output directory.
Megatron-LM/tests/functional_tests/test_cases/gpt/gpt_dynamic_inference_tp1_pp1_583m_cuda_graphs_validation/cuda_graphs.py:16
↓ 1 callersFunctioncli
Megatron Core Distributed Checkpoint Inspector/Editor
Megatron-LM/tools/checkpoint/checkpoint_inspector.py:64
↓ 1 callersFunctionclip
( weights: torch.Tensor, loss_mask: torch.Tensor, metrics: dict[str, list[torch.Tensor]], lowe
slime/examples/train_infer_mismatch_helper/mis.py:116
↓ 1 callersFunctionclip_qk
Clip the QK attention logits to the threshold, recommended for Muon optimizer. Args: model: The model to clip the QK attention logit
Megatron-LM/megatron/core/optimizer/qk_clip.py:8
↓ 1 callersMethodclip_qk
QK Clipping is a technique to clip the query and key attention logits to prevent the attention logits from exploding.
Megatron-LM/megatron/core/transformer/attention.py:1042
↓ 1 callersFunctionclone_tensors_in_struct
Copy src to pre-existing tensors in tgt.
Megatron-LM/megatron/core/full_cuda_graph.py:33
↓ 1 callersMethodclose
(self, lease_id: str)
swe-rl/server/swe_env_pool_server.py:174
↓ 1 callersMethodclose
Terminate the async caller at exit of an application or some termination conditions
Megatron-LM/megatron/core/dist_checkpointing/strategies/async_utils.py:168
↓ 1 callersMethodclose
For TemporalAsyncCaller, this method is called explictly in `is_current_async_calls_done` This method make sure the TemporalAsyncCaller termi
Megatron-LM/megatron/core/dist_checkpointing/strategies/async_utils.py:255
↓ 1 callersMethodclose
Wait on the left async requests and terminate the PersistentAsyncCaller Signals the PersistentAsyncCaller by sending a 'DONE' message to make
Megatron-LM/megatron/core/dist_checkpointing/strategies/async_utils.py:414
↓ 1 callersMethodclose
Finalize all calls upon closing. Args: abort (bool, optional): Default to False. Needs to be manually set to true when
Megatron-LM/megatron/core/dist_checkpointing/strategies/async_utils.py:593
↓ 1 callersMethodclose
No resources to release.
slime/examples/geo3k_vlm_multi_turn/env_geo3k.py:53
↓ 1 callersMethodclose_lease
(self, lease_id: str)
gui-rl/env_pool_server.py:265
↓ 1 callersMethodclose_run
(self, run_lease_id: str)
terminal-rl/remote/pool_server.py:280
↓ 1 callersMethodcode_from_dtype
Get the code from the dtype Args: value (Type[numpy.number]): The dtype Returns: int: The code
Megatron-LM/megatron/core/datasets/indexed_dataset.py:62
↓ 1 callersFunctioncollect_context
(snippet: str, doc: str)
slime/examples/search-r1/google_search_server.py:73
↓ 1 callersMethodcollect_step_results
( self, pending_tasks: list[tuple[int, asyncio.Task]], )
gui-rl/agents/qwen3vl_reward_agent.py:398
↓ 1 callersFunctioncollect_stream
(prompt, request_id, stream_generator)
Megatron-LM/examples/inference/gpt/gpt_static_inference.py:113
↓ 1 callersFunctioncollect_train_test_metrics
( logs_dir: str, train_iters: str, output_path: str, is_convergence_test: bool, is_second_
Megatron-LM/tests/functional_tests/python_test_utils/get_test_results_from_tensorboard_logs.py:32
↓ 1 callersMethodcombine
Append the samples from multiple instances into one object.
Megatron-LM/megatron/core/rerun_state_machine.py:1124
↓ 1 callersMethodcombine
( self, hidden_states: torch.Tensor, async_finish: bool = False, allocate_on_c
Megatron-LM/megatron/core/transformer/moe/token_dispatcher.py:1237
↓ 1 callersFunctioncombine_all_res
(acc_dict)
Megatron-LM/examples/multimodal/evaluation/evaluate_video_mvbench.py:83
↓ 1 callersMethodcombine_embeddings
Combine multiple embeddings from different encoders by concatenation. This method is used for combining encoder outputs before input projecti
Megatron-LM/megatron/core/models/mimo/submodules/vision.py:112
↓ 1 callersMethodcombine_embeddings
Combine embeddings from different encoders.
Megatron-LM/megatron/core/models/mimo/submodules/audio.py:94
↓ 1 callersFunctioncombine_tp_tensors
(params, key, dim, tensors)
Megatron-LM/tools/checkpoint/hybrid_conversion.py:59
↓ 1 callersFunctioncombined_1f1b_schedule_for_interleaved_pipelining
Helper method to run combined forward and backward step for A2A communication hiding. This method merges the functionality of `forward_step_helper
Megatron-LM/megatron/core/pipeline_parallel/combined_1f1b.py:111
↓ 1 callersFunctioncombined_1f1b_schedule_for_no_pipelining
Scheduler for 1f1b with no pipelining. This function schedules micro-batches in a way that the forward pass of Transformer layers for one mic
Megatron-LM/megatron/core/pipeline_parallel/combined_1f1b.py:18
↓ 1 callersFunctioncompare_elements
(elem1, elem2)
gui-rl/desktop_env/evaluators/metrics/chrome.py:381
↓ 1 callersFunctioncompare_helpers
(pipeline_parallel_size, num_microbatches, num_model_chunks)
Megatron-LM/tests/unit_tests/pipeline_parallel/test_helpers.py:1
↓ 1 callersFunctioncompare_image_text
(image_path, rule)
gui-rl/desktop_env/evaluators/metrics/docs.py:369
↓ 1 callersFunctioncompare_pdf_content
(content1, content2, text_similarity_threshold)
gui-rl/desktop_env/evaluators/metrics/vscode.py:114
↓ 1 callersFunctioncompile_allocator
Attempt to compile UVM allocator.
Megatron-LM/megatron/core/inference/unified_memory.py:74
↓ 1 callersFunctioncompose_up_no_build
( terminal: Terminal, *, timeout: float, container_name: str, logger: logging.Logger | Non
terminal-rl/remote/docker_compose_utils.py:248
↓ 1 callersFunctioncompression_ratio
( data: str | bytes, *, encoding: str = "utf-8", algorithm: Literal["zlib", "gzip", "bz2", "lz
slime/slime/utils/metric_utils.py:69
↓ 1 callersFunctioncompute_activation_memory
(args, num_microbatches, verbose=False)
Megatron-LM/megatron/training/theoretical_memory_usage.py:189
↓ 1 callersFunctioncompute_activation_memory_without_sp
Compute activation memory without sequence parallelism
Megatron-LM/megatron/training/theoretical_memory_usage.py:263
↓ 1 callersFunctioncompute_all_acc
(result_list)
Megatron-LM/examples/multimodal/evaluation/evaluate_video_phys_game_bench.py:52
↓ 1 callersMethodcompute_all_pairs
(guesses: List[str], answers: List[str], n=1)
Megatron-LM/tools/retro/text_generation/metrics.py:66
↓ 1 callersFunctioncompute_average_score
Compute average, excluding failed evaluations (-1.0).
openclaw-rl/oel/eval/personalization_evaluator.py:214
↓ 1 callersMethodcompute_default_rope_parameters
(config)
slime/slime_plugins/megatron_bridge/qwen35_vl.py:238
↓ 1 callersMethodcompute_each_pair
(guess: str, answer: str, n=1)
Megatron-LM/tools/retro/text_generation/metrics.py:53
↓ 1 callersFunctioncompute_entropy_from_logits
(logits: torch.Tensor, process_group)
slime/slime/utils/ppo_utils.py:197
↓ 1 callersFunctioncompute_f1_score
Evaluating F1 Score
Megatron-LM/tools/retro/text_generation/evaluate.py:34
↓ 1 callersFunctioncompute_feature_bank
(model)
Megatron-LM/megatron/legacy/model/vision/knn_monitor.py:38
↓ 1 callersFunctioncompute_gpqa_reward
Rule-based scorer for GPQA-style multiple-choice evaluation.
slime/slime/rollout/rm_hub/gpqa.py:54
↓ 1 callersFunctioncompute_group_stats
Add group-based rollout stats for logging. Args: rollouts: Rollouts to generate the stats for. Each inner list is a group (as in GRPO gro
Megatron-LM/megatron/rl/rl_utils.py:743
↓ 1 callersFunctioncompute_grpo_advantages
GRPO-style advantage: broadcast reward as advantage (no normalization). Matches OpenClaw-RL's --disable-rewards-normalization behavior.
openclaw-tinker/data_formatter.py:212
↓ 1 callersFunctioncompute_ifbench_reward
Score a model response using the official IFBench rules.
slime/slime/rollout/rm_hub/ifbench.py:154
↓ 1 callersFunctioncompute_intermediate_size
(n, ffn_dim_multiplier=1, multiple_of=256)
Megatron-LM/tools/checkpoint/loader_llama_mistral.py:61
↓ 1 callersFunctioncompute_mathvista_accuracy
Compute MathVista accuracy.
Megatron-LM/examples/multimodal/evaluation/evaluate_mathvista.py:74
↓ 1 callersFunctioncompute_ocrbench_score
Compute OCRBench score.
Megatron-LM/examples/multimodal/evaluation/evaluate_ocrbench.py:33
↓ 1 callersMethodcompute_prm_teacher_log_probs
Compute log-probs on the dedicated PRM teacher GPU using hint-enhanced tokens. Two paths share this entry point: * Single-candidate
slime/slime/backends/megatron_utils/actor.py:632
↓ 1 callersFunctioncompute_routing_scores_for_aux_loss
Compute routing scores based on the score function. Args: logits (torch.Tensor): The logits tensor after gating, shape: [num_tokens, num_
Megatron-LM/megatron/core/transformer/moe/moe_utils.py:634
↓ 1 callersFunctioncompute_score
The scoring function for countdown task. Args: solution_str: the solution text ground_truth: dictionary containing target number
Megatron-LM/examples/rl/environments/countdown/countdown.py:70
↓ 1 callersFunctioncompute_score
Compute the reward score for a solution. Args: solution_str: The solution string ground_truth: The ground truth answer co
slime/slime/rollout/rm_hub/math_dapo_utils.py:262
↓ 1 callersFunctioncompute_score_em
The scoring function for exact match (EM). Args: solution_str: the solution text ground_truth: the ground truth method: t
slime/examples/search-r1/qa_em_format.py:156
↓ 1 callersFunctioncompute_throughputs_and_append_to_progress_log
(iteration, num_floating_point_operations_so_far)
Megatron-LM/megatron/training/training.py:1825
↓ 1 callersMethodcompute_true_vocab_size
Determine the 'true' (non-padded) vocab size.
Megatron-LM/tools/checkpoint/loader_base.py:157
↓ 1 callersFunctioncompute_weight_and_optimizer_memory
(args, verbose=False)
Megatron-LM/megatron/training/theoretical_memory_usage.py:12
↓ 1 callersFunctionconfigure_nvtx_profiling
Configure NVTX range profiling to be enabled or disabled. Args: enabled (bool): Whether to enable NVTX range profiling
Megatron-LM/megatron/core/utils.py:2110
↓ 1 callersFunctionconfigure_vm_network
(vm_name, interface_name=None)
gui-rl/desktop_env/providers/virtualbox/manager.py:141
↓ 1 callersMethodconnect_rollout_engines
Split colocated/distributed engines. Global source rank (DP=TP=PP=0) creates NCCL for distributed. Map ranks to colocated IPC engines
slime/slime/backends/megatron_utils/update_weight/update_weight_from_tensor.py:66
↓ 1 callersMethodconnect_rollout_engines
( self, rollout_engines: Sequence[ActorHandle], rollout_engine_lock: ActorHandle | Non
slime/slime/backends/fsdp_utils/update_weight_utils.py:40
↓ 1 callersMethodconsume_completion
( self, chat_completion: Any )
terminal-rl/rollout_agent.py:21
↓ 1 callersMethodcontinuous_worker_loop
(self)
openclaw-rl/openclaw_rollout.py:44
↓ 1 callersMethodcontinuous_worker_loop
(self)
openclaw-rl/oel/openclaw_oel_rollout.py:57
↓ 1 callersMethodcontinuous_worker_loop
(self)
openclaw-opd/openclaw_opd_rollout.py:48
↓ 1 callersMethodcontinuous_worker_loop
(self)
openclaw-combine/openclaw_combine_rollout.py:49
↓ 1 callersMethodcontinuous_worker_loop
(self)
openclaw-combine/openclaw_combine_select_rollout.py:63
↓ 1 callersFunctionconvert
(output_path, tensor_parallel_size, use_te)
Megatron-LM/examples/multimodal/model_converter/siglip_converter.py:8
↓ 1 callersFunctionconvert
Convert InternViT HF checkpoint to mcore.
Megatron-LM/examples/multimodal/model_converter/internvit_converter.py:8
↓ 1 callersFunctionconvert
(download_root, output_path, tensor_parallel_size, use_te)
Megatron-LM/examples/multimodal/model_converter/clip_converter.py:10
↓ 1 callersFunctionconvert
(output_path, tensor_parallel_size, use_te, model_type, version)
Megatron-LM/examples/multimodal/model_converter/radio_converter.py:279
↓ 1 callersFunctionconvert_checkpoint
Convert a Megatron Core Distributed Checkpoint from torch_dist to standard fsdp_dtensor format.
Megatron-LM/tools/checkpoint/checkpoint_inspector.py:336
↓ 1 callersFunctionconvert_deepseekv3_to_hf
(args, name, param)
slime/slime/backends/megatron_utils/megatron_to_hf/deepseekv3.py:6
↓ 1 callersFunctionconvert_file
Convert a single safetensors file from quantized format to BF16 (GPU accelerated).
slime/tools/convert_k2_thinking_int4_to_bf16.py:99
↓ 1 callersFunctionconvert_fp8
(input_path, output_path, strategy, block_size=None, max_workers=4)
slime/tools/convert_hf_to_fp8.py:157
↓ 1 callersFunctionconvert_glm4_to_hf
(args, name, param)
slime/slime/backends/megatron_utils/megatron_to_hf/glm4.py:5
↓ 1 callersFunctionconvert_glm4moe_to_hf
(args, name, param)
slime/slime/backends/megatron_utils/megatron_to_hf/glm4moe.py:6
↓ 1 callersFunctionconvert_int4
(input_path, output_path, group_size, is_symmetric, ignore_rules, max_workers)
slime/tools/convert_hf_to_int4_direct.py:205
↓ 1 callersFunctionconvert_llama_to_hf
(args, name, param)
slime/slime/backends/megatron_utils/megatron_to_hf/llama.py:5
↓ 1 callersFunctionconvert_llava_video_to_wds
Convert a LLaVA-Video dataset (keys: video, conversations, data_source) to WebDataset format. The function walks through every *.json / *.jsonl a
Megatron-LM/examples/mimo/data/prepare_video_llava_data.py:23
↓ 1 callersFunctionconvert_mimo_mtp_param
Convert MTP layer parameters from Megatron to HuggingFace format. MTP layers in MiMo contain: - LayerNorms (token_layernorm, hidden_laye
slime/slime/backends/megatron_utils/megatron_to_hf/mimo.py:19
↓ 1 callersFunctionconvert_mimo_to_hf
Convert MiMo model parameters from Megatron to HuggingFace format. MiMo extends Qwen2 with MTP (Multi-Token Prediction) layers.
slime/slime/backends/megatron_utils/megatron_to_hf/mimo.py:6
↓ 1 callersFunctionconvert_point_format
(point, image_ele: dict, *, src_format: str, tgt_format: str)
gui-rl/agents/utils/qwen_vl_utils.py:222
↓ 1 callersFunctionconvert_qwen3moe_to_hf
(args, name, param)
slime/slime/backends/megatron_utils/megatron_to_hf/qwen3moe.py:6
↓ 1 callersFunctionconvert_qwen3vl_to_hf
(args, name, param)
slime/slime/backends/megatron_utils/megatron_to_hf/qwen3_vl.py:5
↓ 1 callersFunctionconvert_radio_g
(output_path, tensor_parallel_size, use_te, version)
Megatron-LM/examples/multimodal/model_converter/radio_converter.py:127
↓ 1 callersFunctionconvert_radio_h
(output_path, tensor_parallel_size, use_te, version)
Megatron-LM/examples/multimodal/model_converter/radio_converter.py:7
↓ 1 callersFunctionconvert_role
(role)
toolcall-rl/sft_data_processing.py:9
↓ 1 callersFunctionconvert_schedule_table_to_order
Convert a tunable schedule lookup table to the te.make_graphed_callables() accepted order format. For example, the tunable schedule table for PP2
Megatron-LM/megatron/core/pipeline_parallel/schedules.py:820
↓ 1 callersFunctionconvert_tasks
Convert terminal bench tasks to parquet format for VERL training. Args: tasks_dir: Directory containing terminal bench tasks (or train ta
terminal-rl/data_utils/convert_task_to_dataset.py:17
↓ 1 callersFunctionconvert_to_coco_format
Convert input files to COCO compatible format.
Megatron-LM/examples/multimodal/evaluation/evaluate_coco.py:10
← previousnext →3,601–3,700 of 10,561, ranked by callers