MCPcopy Create free account

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

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

↓ 2 callersFunction_yarn_get_concentration_factor_from_config
(config: TransformerConfig)
Megatron-LM/megatron/core/models/common/embeddings/yarn_rotary_pos_embedding.py:275
↓ 2 callersMethodaction_last_step
(self)
swe-rl/mini-swe-agent/src/minisweagent/agents/interactive_textual.py:436
↓ 2 callersMethodactivation_func_with_probs
(x, probs)
Megatron-LM/megatron/core/transformer/moe/experts.py:112
↓ 2 callersMethodactive_time
Calculates the cumulative duration for which the timer has been active
Megatron-LM/megatron/core/timers.py:201
↓ 2 callersMethodadd
Track the rng state.
Megatron-LM/megatron/core/extensions/transformer_engine.py:2017
↓ 2 callersMethodadd
(self, meta_info: dict)
slime/slime/utils/types.py:64
↓ 2 callersMethodadd
(self, idx: int, val: int)
slime/slime/utils/seqlen_balancing.py:28
↓ 2 callersMethodadd_event_add
Add 'add' event.
Megatron-LM/megatron/core/inference/inference_request.py:337
↓ 2 callersMethodadd_event_error_nontransient
Add non-transient error event.
Megatron-LM/megatron/core/inference/inference_request.py:357
↓ 2 callersMethodadd_index
Add an entire IndexedDataset to the dataset Args: path_prefix (str): The index (.idx) and data (.bin) prefix
Megatron-LM/megatron/core/datasets/indexed_dataset.py:895
↓ 2 callersFunctionadd_prefix
(t)
Megatron-LM/megatron/core/dist_checkpointing/utils.py:176
↓ 2 callersMethodadd_request
Adds a request to the scheduler and returns the request ID. Args: prompt (str): A prompt string add_BOS (boo
Megatron-LM/megatron/core/inference/engines/static_engine.py:135
↓ 2 callersMethodadd_samples
Add samples to the data source
slime/slime/rollout/data_source.py:25
↓ 2 callersMethodadd_special_tokens
Adds a dictionary of special tokens (eos, pad, cls...). Tokens are only added if they are not already in the vocabulary.
Megatron-LM/megatron/core/tokenizers/text/text_tokenizer.py:130
↓ 2 callersFunctionadd_to_index
Entry point for adding to the index. We select whether to add to a new index, or validate an existing index. Args: config (RetroPrep
Megatron-LM/megatron/core/datasets/retro/index/build.py:299
↓ 2 callersFunctionaggressive_cleanup_memory
More aggressive memory cleanup
toolcall-rl/tool_sandbox.py:52
↓ 2 callersFunctionai2d_eval
Run AI2D evaluation.
Megatron-LM/examples/multimodal/evaluation/evaluate_ai2d.py:38
↓ 2 callersFunctionall_gather_param
All-gather TP-sharded param to full tensor. expert_bias→param, non-TP/duplicated→param.data. Uses expert-TP for ".experts.", else regular-TP.
slime/slime/backends/megatron_utils/update_weight/common.py:50
↓ 2 callersMethodallocate
(self, image: str, instance_id: str = "")
swe-rl/swe_env_client.py:62
↓ 2 callersMethodallocate_all_tensors
Allocate GPU state. This method is used for both 1) initial allocation, and 2) resuming the GPU state after a suspend. Args:
Megatron-LM/megatron/core/inference/contexts/dynamic_context.py:483
↓ 2 callersMethodallocate_slot
Allocates a new slot for a request in the Mamba state buffers. Returns: int: The index of the allocated slot.
Megatron-LM/megatron/core/inference/contexts/attention_context/mamba_metadata.py:87
↓ 2 callersFunctionanalyzePath
* Analyze current pathname to figure out repo root + language segment pattern. * Supports patterns: * /en/… (language as first s
slime/docs/_static/js/lang-toggle.js:18
↓ 2 callersFunctionapply_fsdp2
Apply FSDP v2 to the model. Args: model: The model to wrap with FSDP mesh: Optional DeviceMesh for FSDP. If None, uses all ranks.
slime/slime/backends/fsdp_utils/actor.py:979
↓ 2 callersFunctionapply_nd_flattened_tensors_reformulation
Applies N-D reformulation to a given sharded state dict. After applying the method and loading the reformulated state dict, the `restore_nd_f
Megatron-LM/megatron/core/dist_checkpointing/strategies/resharding.py:96
↓ 2 callersFunctionapply_prefix_mapping
Replaces prefixes *only in keys matching* with one of prefixes in the map. Args: sharded_state_dict (ShardedStateDict): sharded state dic
Megatron-LM/megatron/core/dist_checkpointing/utils.py:210
↓ 2 callersMethodapply_rotary_emb_key
Apply rotary embedding to key tensor. Args: key (Tensor): Key tensor. key_emb (Tensor): Key rotary embeddings.
Megatron-LM/megatron/core/inference/contexts/dynamic_context.py:925
↓ 2 callersMethodapply_rotary_emb_query
Apply rotary embedding to query tensor. Args: query (Tensor): Query tensor. query_emb (Tensor): Query rotary embeddin
Megatron-LM/megatron/core/inference/contexts/dynamic_context.py:891
↓ 2 callersMethodapply_saving_parallelization
Distributes the save across ranks by exchanging metadata. Exchanges metadata from the state dict and computes the uniform (as close a
Megatron-LM/megatron/core/dist_checkpointing/strategies/fully_parallel.py:100
↓ 2 callersFunctionare_lists_equal
(list1, list2, comparison_func)
gui-rl/desktop_env/evaluators/metrics/utils.py:768
↓ 2 callersMethodasync_bucket_gather
All-gather the bucket and set the items.
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:3592
↓ 2 callersMethodasync_compute_student_topk
Run the student-side old_actor pass and return rank-0's student top-K indices. Used by ``--distill-subset-mode student``: the outer loop ship
slime/slime/ray/actor_group.py:152
↓ 2 callersMethodasync_gather_at_indices
Run a PRM-teacher forward gathering log-probs at provided indices.
slime/slime/ray/actor_group.py:161
↓ 2 callersMethodasync_generate_output_tokens_dynamic_batch
Forward step the model and update the inference context. Args: skip_bookkeeping (Optional[bool]): If true, skip the context bookk
Megatron-LM/megatron/core/inference/text_generation_controllers/text_generation_controller.py:880
↓ 2 callersFunctionasync_rm
(args, sample: Sample, **kwargs)
slime/slime/rollout/rm_hub/__init__.py:30
↓ 2 callersMethodasync_save
Perform preparation and return an AsyncRequest to the external caller. Args: sharded_state_dict (ShardedStateDict): sharded state
Megatron-LM/megatron/core/dist_checkpointing/strategies/base.py:206
↓ 2 callersFunctionavailable_memory
()
slime/slime/utils/memory_utils.py:18
↓ 2 callersMethodbackward
Backward pass for scaled softmax (no mask). Args: ctx: Autograd context containing saved tensors from forward. output
Megatron-LM/megatron/core/fusions/fused_softmax.py:136
↓ 2 callersMethodbackward
Schedule node backward
Megatron-LM/megatron/core/pipeline_parallel/utils.py:242
↓ 2 callersMethodbackward_dw
Execute weight update operations
Megatron-LM/megatron/core/transformer/attention.py:1265
↓ 2 callersFunctionbackward_step_helper_postprocess
Postprocess for backward_step_helper
Megatron-LM/megatron/core/pipeline_parallel/schedules.py:1293
↓ 2 callersFunctionbackward_step_helper_preprocess
Preprocess for backward_step_helper
Megatron-LM/megatron/core/pipeline_parallel/schedules.py:1274
↓ 2 callersMethodbase_generate
(self, request: InferenceRequest)
Megatron-LM/megatron/rl/inference/inference_interface.py:44
↓ 2 callersFunctionbaseline_get_model_chunk_id
Helper method to get the model chunk ID given the iteration number.
Megatron-LM/tests/unit_tests/pipeline_parallel/test_helpers.py:5
↓ 2 callersMethodbbox_coord_to_label
Format bbox coordinates as text.
Megatron-LM/examples/multimodal/dataset_helpers.py:758
↓ 2 callersFunctionbias_geglu_impl
Implementation of biased GEGLU that handles different input shapes. This function reshapes the input if necessary, applies the GEGLU activation
Megatron-LM/megatron/core/fusions/fused_bias_geglu.py:153
↓ 2 callersFunctionbias_gelu
(bias, y)
Megatron-LM/megatron/legacy/model/fused_bias_gelu.py:16
↓ 2 callersFunctionbias_swiglu
Performs SwiGLU activation with bias addition. Args: y (torch.Tensor): Input tensor. bias (torch.Tensor): Bias tensor to be added
Megatron-LM/megatron/core/fusions/fused_bias_swiglu.py:30
↓ 2 callersFunctionbiencoder_model_provider
Build the model.
Megatron-LM/megatron/legacy/model/biencoder_model.py:37
↓ 2 callersFunctionbit_same
(a, b)
Megatron-LM/tests/unit_tests/a2a_overlap/utils.py:139
↓ 2 callersFunctionbroadcast_float_list
Broadcast a list of float values. Args: data_parallel (bool): Broadcast across a single data parallel model replica.
Megatron-LM/megatron/core/inference/communication_utils.py:202
↓ 2 callersFunctionbroadcast_list
Broadcast a list of values with a given type. Args: data_parallel (bool): Broadcast across a single data parallel model replica.
Megatron-LM/megatron/core/inference/communication_utils.py:169
↓ 2 callersMethodbroadcast_params
Syncs parameters across all DP ranks.
Megatron-LM/megatron/core/distributed/data_parallel_base.py:67
↓ 2 callersMethodbuild_and_save_engine
Method to build the TRTLLM Engine This method uses the TRTLLMEngineBuilder to build and save the engine to engine dir Args:
Megatron-LM/megatron/core/export/trtllm/trtllm_helper.py:532
↓ 2 callersMethodbuild_config_from_dict
Builds an MLPConfig object from a dictionary and a hidden size. Args: block_config_dict (dict): The dictionary containin
Megatron-LM/megatron/core/transformer/heterogeneous/heterogeneous_config.py:74
↓ 2 callersFunctionbuild_datasets
(idir, odir, extra_args=[])
Megatron-LM/tests/unit_tests/data/test_preprocess_data.py:54
↓ 2 callersFunctionbuild_datum
Build a tinker.Datum from token IDs with appropriate weights. Weights are 0 for prompt tokens and 1 for response tokens.
openclaw-fireworks/fireworks_loss.py:110
↓ 2 callersFunctionbuild_distributed_environment
Helper function to build a distributed environment for testing Megatron-FSDP. Order of dimensions is (DP_OUTER, DP_SHARD, CP, TP).
Megatron-LM/tests/unit_tests/distributed/fsdp/test_mfsdp_fully_shard.py:168
↓ 2 callersFunctionbuild_dynamic_engine_setup_prefix
Returns a compact, pipe-separated summary of the dynamic-batching setup. Example output: `dynamic | cg True | prompts: synth(16 256), n
Megatron-LM/examples/inference/gpt/utils.py:371
↓ 2 callersFunctionbuild_gpt_chunk_datasets_from_gpt_datasets
Get train, valid, test GPT chunk datasets. Args: project_dir (str): Retro project dir. gpt_datasets (dict): Mapping of 'train', '
Megatron-LM/megatron/core/datasets/retro/query/gpt_chunk_dataset.py:79
↓ 2 callersFunctionbuild_hint_judge_messages
Hint judge prompt — used by OPD and Combined methods.
openclaw-tinker/scorers.py:85
↓ 2 callersFunctionbuild_layer_callables
Builds the callable functions(forward and dw) for the given layer. For now, 1f1b overlap only support TransformerLayer and MultiTokenPredicti
Megatron-LM/megatron/core/models/gpt/fine_grained_callables.py:599
↓ 2 callersFunctionbuild_model
()
Megatron-LM/megatron/training/training.py:971
↓ 2 callersMethodbuild_policy_messages
Build policy messages with a bounded recent-history window. Keep the same retention rule as the working Qwen3VL GUI path: on
gui-rl/agents/qwen35_agent.py:283
↓ 2 callersMethodbuild_train_system_message
(self)
gui-rl/agents/qwen35_agent.py:265
↓ 2 callersFunctionbuild_train_valid_test_datasets
Build pretraining datasets.
Megatron-LM/megatron/training/training.py:2960
↓ 2 callersFunctionbuild_transformer_layer_callables
Create callables for transformer layer nodes. Divides the transformer layer's operations into a sequence of smaller, independent functions. Th
Megatron-LM/megatron/core/models/gpt/fine_grained_callables.py:317
↓ 2 callersMethodbulk_reload_group
Bulk reload group.
Megatron-LM/megatron/core/pipeline_parallel/fine_grained_activation_offload.py:354
↓ 2 callersFunctioncalc_params_l2_norm
Calculate l2 norm of parameters
Megatron-LM/megatron/training/utils.py:48
↓ 2 callersFunctioncalculate_contrast
Calculate the contrast of an image as the standard deviation of the pixel values.
gui-rl/desktop_env/evaluators/metrics/gimp.py:184
↓ 2 callersMethodcalculate_cross_entropy_loss
Calculates cross entropy loss.
Megatron-LM/megatron/core/tensor_parallel/cross_entropy.py:71
↓ 2 callersMethodcalculate_gradients
Calculates gradients.
Megatron-LM/megatron/core/tensor_parallel/cross_entropy.py:104
↓ 2 callersFunctioncalculate_image_sharpness
(image_path)
gui-rl/desktop_env/evaluators/metrics/gimp.py:191
↓ 2 callersFunctioncalculate_ins_level_acc
Calculate the instruction level accuracy for given Subject results
Megatron-LM/examples/multimodal/evaluation/mmmu_utils.py:393
↓ 2 callersFunctioncalculate_layer_flops
(args, seqlen, hidden_size, num_attention_heads, num_query_groups, ffn_hidden_size)
slime/slime/utils/flops_utils.py:57
↓ 2 callersMethodcalculate_logits_max
Calculates logits_max.
Megatron-LM/megatron/core/tensor_parallel/cross_entropy.py:23
↓ 2 callersFunctioncalculate_num_audio_tokens
(audio_tensor, model_name)
Megatron-LM/examples/mimo/data/utils/calculate_audio_tokens.py:44
↓ 2 callersMethodcalculate_predicted_logits
Calculates predicted logits.
Megatron-LM/megatron/core/tensor_parallel/cross_entropy.py:35
↓ 2 callersFunctioncalculate_saturation
(image)
gui-rl/desktop_env/evaluators/metrics/gimp.py:63
↓ 2 callersMethodcall_ddp_preforward_hook
Call any DDP pre-forward hooks which are used to launch async data parallel param gather. Any other pre-forward hooks are not allowed.
Megatron-LM/megatron/core/transformer/cuda_graphs.py:1144
↓ 2 callersFunctioncall_gpt
Call Azure GPT model. If conversation is provided, use it instead of system+user.
openclaw-rl/oel/eval/gsm8k_personal_agent.py:188
↓ 2 callersFunctioncall_rollout_fn
(fn, *args, evaluation: bool, **kwargs)
slime/slime/rollout/base_types.py:19
↓ 2 callersFunctionceil_by_factor
返回大于等于 number 的且能被 factor 整除的整数
gui-rl/agents/utils/qwen_vl_utils.py:9
↓ 2 callersFunctionceil_div
(x: int, y: int)
Megatron-LM/megatron/core/extensions/transformer_engine.py:1165
↓ 2 callersFunctionceil_div
Perform ceiling division of two integers. Args: x: the dividend. y: the divisor. Returns: The resul
slime/slime/backends/megatron_utils/kernels/fp8_kernel.py:10
↓ 2 callersFunctionceildiv
(a, b)
slime/tools/convert_hf_to_fp8.py:35
↓ 2 callersFunctionchartqa_eval
Run ChartQA evaluation.
Megatron-LM/examples/multimodal/evaluation/evaluate_chartqa.py:35
↓ 2 callersFunctioncheck_adlr_autoresume_termination
Check for autoresume signal and exit if it is received.
Megatron-LM/megatron/training/utils.py:312
↓ 2 callersMethodcheck_availability
Check if the request can be added to the context.
Megatron-LM/megatron/core/inference/contexts/dynamic_context.py:1396
↓ 2 callersMethodcheck_backend_compatibility
Verifies if this strategy is compatible with `loaded_backend`.
Megatron-LM/megatron/core/dist_checkpointing/strategies/base.py:81
↓ 2 callersFunctioncheck_equal
Check if two inputs are equal, used for checking checkpointing.
Megatron-LM/tests/unit_tests/dist_checkpointing/test_layer_wise_optimizer.py:37
↓ 2 callersMethodcheck_equal_dp_zero_state
( self, dp_zero_state_A, dp_zero_state_B, same_dp_group, raise_if_different=False )
Megatron-LM/tests/unit_tests/dist_checkpointing/test_optimizer.py:801
↓ 2 callersFunctioncheck_gradient
Compute (rel_error, bound, ok) for layer l. - If dtype is None, infer from g_ref (or g_hat if needed). # See https://www.arxiv.org/pdf/25
Megatron-LM/tests/functional_tests/python_test_utils/test_optimizer_grads_match.py:80
↓ 2 callersFunctioncheck_mamba_sequence_packing_support
Checks whether `causal_conv1d` and `mamba_ssm` support sequence packing.
Megatron-LM/megatron/core/utils.py:468
↓ 2 callersFunctioncheck_param_hashes_across_dp_replicas
Computes hashes of all parameters in model, all-gathers hashes across DP replicas, and then checks for equality between the locally-computed hashe
Megatron-LM/megatron/core/utils.py:800
↓ 2 callersMethodcheck_version_compatibility
Verifies if this strategy is compatible with `loaded_version`.
Megatron-LM/megatron/core/dist_checkpointing/strategies/base.py:86
↓ 2 callersFunctionchunked_gae
Compute Generalized Advantage Estimation (GAE) using a FlashLinearAttention- inspired algorithm: parallel prefix scan within chunks and recur
slime/slime/utils/ppo_utils.py:506
↓ 2 callersMethodcleanup
Unload model and release Tinker sessions.
openclaw-tinker/trainer.py:293
↓ 2 callersMethodcleanup
(self)
Megatron-LM/tests/unit_tests/fusions/test_fused_linear_cross_entropy.py:573
↓ 2 callersFunctionclear_embedding_activation_buffer
Clear embedding activation buffer.
Megatron-LM/megatron/core/pipeline_parallel/schedules.py:695
↓ 2 callersFunctionclip_grad_by_total_norm_fp32
Clips gradient of an iterable of parameters in fp32 by total norm. Note that the gradients are modified in place. Args: parameters (
Megatron-LM/megatron/core/optimizer/clip_grads.py:138
← previousnext →1,701–1,800 of 10,561, ranked by callers