MCPcopy Create free account

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

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

↓ 1 callersFunctionextract_data_config
Extract data config from dataset. Args: config (RetroPreprocessingConfig): Retro preprocessing config. Returns: The config o
Megatron-LM/megatron/core/datasets/retro/utils.py:67
↓ 1 callersFunctionextract_information_blocks
(text: str)
slime/examples/search-r1/qa_em_format.py:141
↓ 1 callersFunctionextract_main_log_to_string
(logs_path: pathlib.Path)
Megatron-LM/tests/test_utils/python_scripts/launch_jet_workload.py:204
↓ 1 callersFunctionextract_numbers
Exact all forms of numbers from a string with regex.
Megatron-LM/examples/multimodal/evaluation/mmmu_utils.py:243
↓ 1 callersMethodextract_selected_solution_idx
Extracts the selected solution ID from the response.
slime/examples/multi_agent/agent_system.py:150
↓ 1 callersFunctionextract_sharded_tensors
Extract a dict consisting of only ShardedTensor objects from a given state dict with any objects. Args: sharded_state_dict: state dic
Megatron-LM/megatron/core/dist_checkpointing/utils.py:68
↓ 1 callersFunctionextract_solution
Extract the equation from the solution string.
Megatron-LM/examples/rl/environments/countdown/countdown.py:17
↓ 1 callersFunctionextract_solution
Extract the equation from the solution string.
slime/examples/search-r1/qa_em_format.py:126
↓ 1 callersFunctionextract_text_shapes
(shape)
gui-rl/desktop_env/evaluators/metrics/slides.py:169
↓ 1 callersFunctionextract_torchrunlogs_to_string
(logs_path: pathlib.Path)
Megatron-LM/tests/test_utils/python_scripts/launch_jet_workload.py:180
↓ 1 callersFunctionf1_score
(prediction, ground_truth)
slime/slime/rollout/rm_hub/f1.py:24
↓ 1 callersFunctionfake_int4_quantization_ste
(x, group_size)
Megatron-LM/megatron/core/extensions/transformer_engine.py:1213
↓ 1 callersFunctionfetch
(session: aiohttp.ClientSession, url: str, semaphore: asyncio.Semaphore)
slime/examples/search-r1/google_search_server.py:42
↓ 1 callersFunctionfetch_all
(urls: list[str], limit: int = 8)
slime/examples/search-r1/google_search_server.py:63
↓ 1 callersFunctionfetch_github_issue
Fetch GitHub issue text from the URL.
swe-rl/mini-swe-agent/src/minisweagent/run/github_issue.py:22
↓ 1 callersMethodffn_mult_to_intermediate_size
Calculates the intermediate size of the MLP based on the given `ffn_mult` and `hidden_size`. Args: ffn_mult (flo
Megatron-LM/megatron/core/transformer/heterogeneous/heterogeneous_config.py:101
↓ 1 callersMethodfill_in_deferred_sharded_objects
Fill in objects not loaded by current rank with objects from `loaded_objects` map. Args: sharded_state_dict (ShardedStateDict): s
Megatron-LM/megatron/core/dist_checkpointing/strategies/fully_parallel.py:317
↓ 1 callersMethodfill_in_deferred_sharded_tensors
Fill in tensors not loaded by current rank with tensors from `loaded_tensors` map. Args: sharded_state_dict (ShardedStateDict): s
Megatron-LM/megatron/core/dist_checkpointing/strategies/fully_parallel.py:336
↓ 1 callersMethodfill_routing_replay
(self, data_iterator, num_microbatches, rollout_data)
slime/slime/backends/megatron_utils/actor.py:516
↓ 1 callersFunctionfilter_by_environment
(workload_manifests: List[dotdict], environment: str)
Megatron-LM/tests/test_utils/python_scripts/recipe_parser.py:133
↓ 1 callersFunctionfilter_by_model
Returns all workload with matching model.
Megatron-LM/tests/test_utils/python_scripts/recipe_parser.py:168
↓ 1 callersFunctionfilter_by_platform
(workload_manifests: List[dotdict], platform: str)
Megatron-LM/tests/test_utils/python_scripts/recipe_parser.py:151
↓ 1 callersFunctionfilter_by_scope
Returns all workload with matching scope.
Megatron-LM/tests/test_utils/python_scripts/recipe_parser.py:118
↓ 1 callersFunctionfilter_by_tag
Returns all workload with matching tag.
Megatron-LM/tests/test_utils/python_scripts/recipe_parser.py:183
↓ 1 callersFunctionfilter_by_test_case
Returns a workload with matching name. Raises an error if there no or more than a single workload.
Megatron-LM/tests/test_utils/python_scripts/recipe_parser.py:99
↓ 1 callersFunctionfilter_by_test_cases
Returns a workload with matching name. Raises an error if there no or more than a single workload.
Megatron-LM/tests/test_utils/python_scripts/recipe_parser.py:199
↓ 1 callersFunctionfilter_long_prompt
(origin_samples: list[Sample], tokenizer, processor, max_length: int | None)
slime/slime/utils/data.py:81
↓ 1 callersFunctionfilter_unflattened_state_dict
Recursively traverses an unflattened state_dict and collects keys of items that meet the visit_condition. The keys are returned as lists
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/uneven_dtensor.py:202
↓ 1 callersFunctionfinalize_checkpoint
(sample_model, model, params, verbose=False)
Megatron-LM/tools/checkpoint/hybrid_conversion.py:170
↓ 1 callersFunctionfinalize_model_grads
All-reduce all model grads across DP replicas, layernorm grads for sequence parallelism, embedding grads across first and last pipeline stage
Megatron-LM/megatron/core/distributed/finalize_model_grads.py:396
↓ 1 callersMethodfinalize_response
(self, model_response: Any)
terminal-rl/rollout_agent.py:27
↓ 1 callersFunctionfindContainer
()
slime/docs/_static/js/lang-toggle.js:73
↓ 1 callersMethodfind_all_sublist_indices
(self, main_list, sublist)
slime/slime/utils/mask_utils.py:18
↓ 1 callersMethodfind_multiple
Calculates the smallest multiple of `k` greater than or equal to `n`. Args: n (int): The number to find the multiple of.
Megatron-LM/megatron/core/transformer/heterogeneous/heterogeneous_config.py:117
↓ 1 callersFunctionfine_grained_offloading_init_chunk_handler
Initialize the chunk handler, called at the start of a microbatch forward pass.
Megatron-LM/megatron/core/pipeline_parallel/fine_grained_activation_offload.py:566
↓ 1 callersMethodfinish
Finish the checkpointing process. Args: metadata (Metadata): metadata to save results (List[List[WriteResult
Megatron-LM/megatron/core/dist_checkpointing/strategies/filesystem_async.py:475
↓ 1 callersMethodfinish
Close the tensorboard writer
slime/slime/utils/tensorboard_utils.py:59
↓ 1 callersFunctionfinish_mpu_init
()
Megatron-LM/megatron/training/initialize.py:118
↓ 1 callersMethodfinish_param_sync
Finishes param sync communication operation for this bucket. Dispatches next bucket's param sync if available, unless skip_next_bucke
Megatron-LM/megatron/core/distributed/param_and_grad_buffer.py:301
↓ 1 callersMethodfix_query_key_value_ordering
Derives `query`, `key` and `value` tensors from `mixed_qkvz` and `mixed_ba`.
slime/slime_plugins/models/qwen3_next.py:76
↓ 1 callersFunctionflan_format
(system, context, dialogue_turn, template_id=0)
Megatron-LM/tools/retro/sft/dataset_conv.py:317
↓ 1 callersMethodflash_decode
The flash decoding kernel will do the following in a single execution: 1. Compute RoPE embedding with precomputed cos & sin tensors
Megatron-LM/megatron/core/transformer/attention.py:513
↓ 1 callersFunctionflatten
Recursively flatten nested dict into [(key_path, tensor), …].
Megatron-LM/examples/mimo/utils/data_helpers.py:15
↓ 1 callersFunctionflatten_products
Flattens a nested dict of products
Megatron-LM/tests/test_utils/python_scripts/recipe_parser.py:49
↓ 1 callersFunctionflatten_state_dict
Flattens state dict into a single level dict. It's a copy of torch.distributed.checkpoint._nested_dict.flatten_state_dict which also accepts
Megatron-LM/megatron/core/dist_checkpointing/strategies/torch.py:107
↓ 1 callersFunctionflatten_workload
Flattens a workload with products into a list of workloads that don't have products.
Megatron-LM/tests/test_utils/python_scripts/recipe_parser.py:62
↓ 1 callersFunctionfloat16_to_fp32
Convert fp16/bf16 `val` to fp32
Megatron-LM/megatron/legacy/model/module.py:152
↓ 1 callersMethodflush_cache
Flush the cache of the server.
slime/slime/backends/sglang_utils/sglang_engine.py:297
↓ 1 callersMethodfold_responses
( self, responses: list[InferenceResponse], n: int )
Megatron-LM/megatron/rl/inference/inference_interface.py:52
↓ 1 callersFunctionforce_all_tensors_to_non_fp8
Force all tensors in state dict to be non-fp8. Args: sharded_state_dict (ShardedStateDict): sharded state dict.
Megatron-LM/megatron/core/dist_checkpointing/utils.py:236
↓ 1 callersMethodfork
Fork the cuda rng state, perform operations, and exit with the original state.
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/utils.py:341
↓ 1 callersFunctionformat_answer
(answer)
Megatron-LM/tools/retro/sft/dataset_conv.py:177
↓ 1 callersFunctionformat_conversation_with_tools
Format conversation using Jinja2 template with tool support
toolcall-rl/generate_with_retool.py:151
↓ 1 callersFunctionformat_multichoice
(multichoice_options)
Megatron-LM/tools/retro/sft/dataset_conv.py:167
↓ 1 callersFunctionformat_multichoice_answer
Format multi-choice answer.
Megatron-LM/examples/multimodal/dataset_helpers.py:964
↓ 1 callersFunctionformat_multichoice_question
(question, multichoice_options)
Megatron-LM/tools/retro/sft/dataset_conv.py:173
↓ 1 callersFunctionformat_multichoice_question
Format multi-choice question.
Megatron-LM/examples/multimodal/dataset_helpers.py:953
↓ 1 callersMethodformat_observation
(self, observation: dict)
slime/examples/geo3k_vlm_multi_turn/base_env.py:15
↓ 1 callersMethodforward
(ctx, logits, tp_group, tp_world)
openclaw-combine/hint_opd_loss.py:183
↓ 1 callersMethodforward
(self, input_ids, position_ids, attention_mask, retriever_input_ids=None, retr
Megatron-LM/megatron/legacy/model/gpt_model.py:78
↓ 1 callersMethodforward
Forward function of the Mamba model. This function passes the input tensors through the embedding layer, and then the decoder and finally into
Megatron-LM/megatron/core/models/mamba/mamba_model.py:171
↓ 1 callersMethodforward
Forward pass of Yarn Rotary Embedding. Args: max_seq_len (int): Maximum size of sequence offset (int, optional): RoPE
Megatron-LM/megatron/core/models/common/embeddings/yarn_rotary_pos_embedding.py:161
↓ 1 callersMethodforward
Execute the wrapped module in model precision and optionally upcast outputs to fp32. On the first pipeline stage, positional/keyword
Megatron-LM/megatron/core/transformer/module.py:420
↓ 1 callersMethodforward
Forward of ColumnParallelLinearGathered
Megatron-LM/megatron/core/transformer/heterogeneous/linear_replacements.py:99
↓ 1 callersMethodforward
(self, x)
Megatron-LM/tests/unit_tests/test_optimizer.py:47
↓ 1 callersMethodforward
(self, x, use_offload: bool = False)
Megatron-LM/tests/unit_tests/pipeline_parallel/test_fine_grained_activation_offloading.py:41
↓ 1 callersMethodforward
(self, input_features, seq_lengths=None)
Megatron-LM/tests/unit_tests/models/test_mimo_audio_submodules.py:78
↓ 1 callersMethodforward
Run RMSNorm with an option to compute custom statistic.
Megatron-LM/examples/multimodal/nvlm/internvit.py:95
↓ 1 callersMethodforward_all_selfattention
(self, x)
Megatron-LM/megatron/legacy/model/vision/esvit_swin_backbone.py:653
↓ 1 callersMethodforward_features
(self, x)
Megatron-LM/megatron/legacy/model/vision/mit_backbone.py:296
↓ 1 callersMethodforward_fused_softmax
(self, input, mask)
Megatron-LM/megatron/legacy/model/fused_softmax.py:192
↓ 1 callersMethodforward_fused_softmax
Compute softmax using fused CUDA kernels when available. Args: input (torch.Tensor): Attention scores of shape [b, np, sq, sk].
Megatron-LM/megatron/core/fusions/fused_softmax.py:272
↓ 1 callersMethodforward_last_selfattention
(self, x)
Megatron-LM/megatron/legacy/model/vision/esvit_swin_backbone.py:644
↓ 1 callersMethodforward_pass_with_pipeline_parallel_large_input_batch
Utility to carry out forward pass PP models. Runs the forward pass for models which are pipeline parallel. This is more complex than
Megatron-LM/megatron/core/inference/model_inference_wrappers/abstract_model_inference_wrapper.py:261
↓ 1 callersMethodforward_pass_with_pipeline_parallel_small_input_batch
Utility to carry out forward pass for PP models with very small inputs If a model is pipeline parallel, yet, the input global batch is very s
Megatron-LM/megatron/core/inference/model_inference_wrappers/abstract_model_inference_wrapper.py:218
↓ 1 callersMethodforward_pass_without_pipeline_parallel
Utility to carry out simple forward pass for TP or no model parallel models Runs a very simple forward pass for model. Used in the case of m
Megatron-LM/megatron/core/inference/model_inference_wrappers/abstract_model_inference_wrapper.py:199
↓ 1 callersFunctionforward_step
Forward step.
Megatron-LM/pretrain_bert.py:124
↓ 1 callersFunctionforward_step_helper
Helper method to run forward step with model split into chunks
Megatron-LM/megatron/core/pipeline_parallel/schedules.py:1240
↓ 1 callersMethodforward_torch_softmax
(self, input, mask)
Megatron-LM/megatron/legacy/model/fused_softmax.py:210
↓ 1 callersMethodforward_torch_softmax
Fallback PyTorch implementation for masked softmax. Applies optional scaling, constructs a causal or sliding-window mask if needed, a
Megatron-LM/megatron/core/fusions/fused_softmax.py:299
↓ 1 callersMethodforward_with_features
(self, x)
Megatron-LM/megatron/legacy/model/vision/esvit_swin_backbone.py:436
↓ 1 callersFunctionfp32_to_float16
Converts floating-point values from fp32 to fp16. Args: val: The value to convert. Can be a single number, a tuple, or a list. fl
Megatron-LM/megatron/core/transformer/module.py:342
↓ 1 callersFunctionfp8_create_transpose_cache
Create the transpose cache of a FP8 tensor.
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/mixed_precision.py:188
↓ 1 callersFunctionfp8_need_transpose_data_for_meta_device_init
Check if a FP8 tensor needs transpose data, for meta device init scenario.
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/mixed_precision.py:172
↓ 1 callersMethodfree
free a temporary bucket.
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:772
↓ 1 callersMethodfreeze
Freezes the async request, disallowing adding new finalization functions. Returns: AsyncRequest: new async request with all same
Megatron-LM/megatron/core/dist_checkpointing/strategies/async_utils.py:103
↓ 1 callersMethodfreeze
( self, freeze_language_model: bool, freeze_vision_model: bool, freeze_vision_
slime/slime_plugins/megatron_bridge/qwen35_vl.py:693
↓ 1 callersFunctionfreeze_model_params
(model: GPTModel, args: argparse.Namespace)
slime/slime/backends/megatron_utils/model_provider.py:324
↓ 1 callersMethodfrom_dir
Get list of block files, and create map. Args: dir (str): Path to directory containing saved block files. block_size
Megatron-LM/megatron/core/datasets/retro/utils.py:343
↓ 1 callersMethodfrom_string
(cls, tokenizer_name: str)
Megatron-LM/megatron/rl/inference/chat_templates.py:32
↓ 1 callersMethodfrom_yaml_file
Loads recipe from yaml configuration.
Megatron-LM/megatron/core/quantization/quant_config.py:168
↓ 1 callersFunctionfully_shard_model
Fully-shard the model for Megatron-FSDP. Args: Subset of the arguments for fully_shard(). Refer to the docstring for fully_shard().
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/fully_shard.py:60
↓ 1 callersFunctionfully_shard_optimizer
Fully shard the optimizer for Megatron-FSDP. This is an in-place operation on the optimizer instance, which modifies the optimizer to call me
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/fully_shard.py:230
↓ 1 callersFunctionfused_apply_rotary_pos_emb
Apply rotary positional embedding to input tensor T in `sbhd` format.
Megatron-LM/megatron/core/extensions/transformer_engine.py:2102
↓ 1 callersFunctionfused_apply_rotary_pos_emb_thd
Apply rotary positional embedding to input tensor T in `thd` format with CP support.
Megatron-LM/megatron/core/extensions/transformer_engine.py:2123
↓ 1 callersFunctionfused_combine
Perform fused combine operation if deep_ep is available. Args: x: Input tensor group: Process group handl
Megatron-LM/megatron/core/transformer/moe/fused_a2a.py:244
↓ 1 callersFunctionfused_dispatch
Perform fused dispatch operation if deep_ep is available. Args: x: Input tensor [num_tokens, hidden_size] token_indic
Megatron-LM/megatron/core/transformer/moe/fused_a2a.py:212
↓ 1 callersFunctionfused_experts_impl
( hidden_states: torch.Tensor, w1: torch.Tensor, w2: torch.Tensor, topk_weights: torch.Tensor,
slime/slime/backends/fsdp_utils/models/qwen3_moe.py:14
↓ 1 callersMethodgather_indices
Gather tensors and concatinate along the first dimension.
Megatron-LM/megatron/legacy/model/transformer.py:227
↓ 1 callersFunctiongather_log_probs_packed
Gather next-token log probabilities for packed sequences. Parameters: logits: Model logits of shape [B, T, V] or [T, V]. input_id
slime/slime/backends/fsdp_utils/actor.py:875
← previousnext →3,901–4,000 of 10,561, ranked by callers