MCPcopy Create free account

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

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

↓ 1 callersFunctionconvert_to_hf
(model_path, input_base_path, model_size, tokenizer_path)
Megatron-LM/tools/checkpoint/loader_llama_mistral.py:77
↓ 1 callersFunctionconvert_to_ocrbench_v2_format
Convert input files to OCRBenchV2 compatible format.
Megatron-LM/examples/multimodal/evaluation/evaluate_ocrbench_v2.py:10
↓ 1 callersFunctionconvert_to_rdtablebench_format
Convert input files to RDTableBench compatible format.
Megatron-LM/examples/multimodal/evaluation/evaluate_rd_tablebench.py:22
↓ 1 callersMethodconvert_tokens_to_ids
Converts a sequence of tokens into ids using the vocab.
Megatron-LM/megatron/training/tokenizer/gpt2_tokenization.py:250
↓ 1 callersMethodconvert_tokens_to_ids
(self, tokens)
Megatron-LM/megatron/training/tokenizer/tokenizer.py:905
↓ 1 callersFunctioncopy_assets
(origin_hf_dir: str, output_dir: str)
slime/tools/convert_fsdp_to_hf.py:152
↓ 1 callersFunctioncopy_assets
(origin_hf_dir, output_dir)
slime/tools/convert_torch_dist_to_hf.py:148
↓ 1 callersFunctioncopy_tensors_in_struct
Copy src to new tensors.
Megatron-LM/megatron/core/full_cuda_graph.py:19
↓ 1 callersFunctioncore_config_from_args
Builds core config object from namespace args from given dataclass Raises exception if argument missing in args Args: args(SimpleNam
Megatron-LM/megatron/training/yaml_arguments.py:379
↓ 1 callersFunctioncore_gpt_dataset_config_from_args
(args)
Megatron-LM/pretrain_gpt.py:192
↓ 1 callersFunctioncore_gpt_dataset_config_from_args
(args)
Megatron-LM/pretrain_mamba.py:151
↓ 1 callersFunctioncore_transformer_config_from_yaml
(args, transfomer_key = "language_model")
Megatron-LM/megatron/training/yaml_arguments.py:403
↓ 1 callersFunctioncorrect_amax_history_if_needed
Correct the amax history of fp8 tensors when it's necessary (i.e., in TE1.x).
Megatron-LM/megatron/core/fp8_utils.py:494
↓ 1 callersFunctioncosine_scheduler
(base_value, final_value, epochs, niter_per_ep, warmup_epochs=0, start_warmup_value=0)
Megatron-LM/megatron/legacy/model/vision/dino.py:159
↓ 1 callersFunctioncount_parameters_in_layer
(model, layer_name)
Megatron-LM/model_provider.py:70
↓ 1 callersFunctioncount_parameters_in_layer
(model, layer_name)
Megatron-LM/megatron/post_training/model_builder.py:28
↓ 1 callersFunctioncount_unknown_letters_in_expr
(expr: str)
slime/slime/rollout/rm_hub/math_utils.py:323
↓ 1 callersMethodcreate
(args)
slime/slime/utils/profile_utils.py:83
↓ 1 callersFunctioncreateButton
()
slime/docs/_static/js/lang-toggle.js:52
↓ 1 callersMethodcreate_attn_mask
(self, H, W)
Megatron-LM/megatron/legacy/model/vision/esvit_swin_backbone.py:229
↓ 1 callersFunctioncreate_checkpoint
Setup a dummy checkpoint directory.
Megatron-LM/tests/unit_tests/test_checkpointing.py:77
↓ 1 callersMethodcreate_decentralized_global_plan
Nothing to do, just some checks. Args: local_plan (SavePlan): local plan to turn to a global plan (without intera
Megatron-LM/megatron/core/dist_checkpointing/strategies/torch.py:548
↓ 1 callersFunctioncreate_empty_bins
Create empty bins for padding to ensure all ranks have the same number of bins. Args: num_empty_bins: Number of empty bins to create
Megatron-LM/megatron/rl/rl_utils.py:119
↓ 1 callersFunctioncreate_env_client
()
terminal-rl/env_client.py:12
↓ 1 callersFunctioncreate_hybrid_dp_cp_groups
Creates groups required for hybrid DPxCP. Creates a new group for every power of 2 up to the number of DPxCP ranks. Returns a dictionary
Megatron-LM/megatron/core/parallel_state.py:417
↓ 1 callersFunctioncreate_mock_image
Create a simple mock image (all zeros). Args: image_size: Size of the square image Returns: Tensor of shape [3, H, W] w
Megatron-LM/examples/mimo/data/mock.py:15
↓ 1 callersFunctioncreate_openai_adapter
Factory function to create OpenAI compatible tool call adapter Args: tools_info: List of tool information parser_type: Parse
slime/examples/tau-bench/openai_tool_adapter.py:162
↓ 1 callersFunctioncreate_prm_agent
( args: Any, *, tokenizer: Any, task_instruction: str, log_tag: str, )
terminal-rl/rollout_agent.py:124
↓ 1 callersFunctioncreate_reference_mlp
Create a reference MLP with full unsharded weights. Args: hidden_size: Input/output dimension ffn_hidden_size: Hidden dimension
Megatron-LM/tests/unit_tests/transformer/test_transformer_block_custom_pgs.py:101
↓ 1 callersMethodcreate_reference_model
Create a reference model by cloning the current model.
Megatron-LM/tests/unit_tests/test_layer_wise_optimizer.py:133
↓ 1 callersMethodcreate_reminder
Create reminder for PR.
Megatron-LM/tests/test_utils/python_scripts/auto_reminder_github.py:228
↓ 1 callersFunctioncreate_result_dict
(result_list)
Megatron-LM/examples/multimodal/evaluation/evaluate_video_mvbench.py:53
↓ 1 callersFunctioncreate_rollout_agent
( *, agent_type: str, args: Any, tokenizer: Any, sampling_params: Dict[str, Any], mode
terminal-rl/rollout_agent.py:74
↓ 1 callersFunctioncreate_run_id
()
slime/slime/utils/external_utils/command_utils.py:227
↓ 1 callersFunctioncreate_test_args
()
Megatron-LM/tests/unit_tests/test_training.py:13
↓ 1 callersFunctioncreate_test_args
(cp_size, sequence_parallel)
Megatron-LM/tests/unit_tests/models/test_llava_model.py:531
↓ 1 callersMethodcreate_test_args
( self, tp, recipe, sequence_length, micro_batch_size, inferen
Megatron-LM/tests/unit_tests/test_fp8_param.py:108
↓ 1 callersMethodcreate_test_args
( self, cuda_graph_impl, cuda_graph_scope, cuda_graph_warmup_steps, ep_size, **kwargs )
Megatron-LM/tests/unit_tests/transformer/test_cuda_graphs.py:1084
↓ 1 callersFunctioncreate_unified_mempool
Create a unified memory mempool using CUDA managed memory. Returns: (MemPool) Unified memory mempool.
Megatron-LM/megatron/core/inference/unified_memory.py:184
↓ 1 callersFunctioncreate_vm_manager_and_provider
Factory function to get the Virtual Machine Manager and Provider instances based on the provided provider name. Args: provider_n
gui-rl/desktop_env/providers/__init__.py:4
↓ 1 callersFunctioncreate_vm_snapshot
(vm_name, max_retries=20)
gui-rl/desktop_env/providers/virtualbox/manager.py:280
↓ 1 callersFunctioncreate_vm_snapshot
(vm_path, max_retries=20)
gui-rl/desktop_env/providers/vmware/manager.py:270
↓ 1 callersMethodcurr_iter
Return current training/validation iteration.
Megatron-LM/megatron/core/full_cuda_graph.py:192
↓ 1 callersFunctioncustom_backward
Directly call C++ autograd engine. To make the 'deallocate_output_tensor' (above) optimization work, the C++ autograd engine must be called d
Megatron-LM/megatron/core/pipeline_parallel/schedules.py:162
↓ 1 callersFunctioncyclic_iter
(iter)
Megatron-LM/megatron/training/training.py:2934
↓ 1 callersFunctioncyclic_iter
(iter)
Megatron-LM/examples/multimodal/dataloader_provider.py:168
↓ 1 callersMethoddata_read
Read all microbatch inputs from Dataloader and copy to static buffers.
Megatron-LM/megatron/core/full_cuda_graph.py:106
↓ 1 callersFunctiondataset_exists
Check if the dataset exists on object storage Args: path_prefix (str): The prefix to the index (.idx) and data (.bin) files idx_
Megatron-LM/megatron/core/datasets/object_storage_utils.py:220
↓ 1 callersFunctiondatasets_provider
Create multimodal train, validation and test datasets.
Megatron-LM/examples/multimodal/dataloader_provider.py:27
↓ 1 callersMethoddeallocate_all_tensors
Deallocate GPU state. This method is used for suspending the dynamic engine.
Megatron-LM/megatron/core/inference/contexts/dynamic_context.py:618
↓ 1 callersFunctiondecoder_model_with_local_default_spec
LLava decoder local spec.
Megatron-LM/megatron/core/models/multimodal/llava_spec.py:65
↓ 1 callersFunctiondecoder_model_with_local_spec
T5 decoder local spec (uses Megatron-Core components).
Megatron-LM/megatron/core/models/T5/t5_spec.py:155
↓ 1 callersFunctiondecoder_model_with_transformer_engine_default_spec
T5 decoder TE spec (uses Transformer Engine components).
Megatron-LM/megatron/core/models/T5/t5_spec.py:79
↓ 1 callersFunctiondecoder_model_with_transformer_engine_default_spec
LLava decoder TE spec (uses Transformer Engine components).
Megatron-LM/megatron/core/models/multimodal/llava_spec.py:37
↓ 1 callersMethoddeduplicate_chunks
Group tensors by chunk and then pick the tensor with the lowest rank. NOTE: with proper loading overlap, loading from randomized ranks
Megatron-LM/megatron/core/dist_checkpointing/strategies/two_stage.py:189
↓ 1 callersMethoddefault_backward_func
Default backward function
Megatron-LM/megatron/core/pipeline_parallel/utils.py:187
↓ 1 callersMethoddefault_decoder_cross_attention
Cross attention for a standard encoder-decoder model.
Megatron-LM/megatron/legacy/model/transformer.py:958
↓ 1 callersMethoddelete_vm
Delete the registration of VM by path.
gui-rl/desktop_env/providers/base.py:65
↓ 1 callersMethoddeserialize
Construct SamplingParams from a msgpack-compatible dictionary.
Megatron-LM/megatron/core/inference/sampling_params.py:52
↓ 1 callersMethoddeserialize
Deserialize error. Args: obj (dict): Serialized error data. Returns: (ContextOverflowError) Deserialized err
Megatron-LM/megatron/core/inference/contexts/dynamic_context.py:158
↓ 1 callersFunctiondeserialize_tensor
Deserialize tensor from bytes. Args: tensor_bytes (bytes): Byte representation of tensor. Returns: (Tensor) Tensor.
Megatron-LM/megatron/core/inference/inference_request.py:33
↓ 1 callersFunctiondestroy_global_memory_buffer
Sets the global memory buffer to None
Megatron-LM/megatron/core/parallel_state.py:1991
↓ 1 callersMethoddestroy_process_groups
()
slime/slime/utils/reloadable_process_group.py:132
↓ 1 callersFunctiondestroy_rerun_state_machine
Helper function to shut down the rerun machine instance.
Megatron-LM/megatron/core/rerun_state_machine.py:1289
↓ 1 callersFunctiondetokenize
Detokenizes the given prompt.
Megatron-LM/megatron/core/inference/text_generation_server/endpoints/completions.py:24
↓ 1 callersMethoddetokenize
Convert embedding ids to text Args: ids (numpy.ndarray): The ids to convert Returns: str: The converted text
Megatron-LM/megatron/core/datasets/megatron_tokenizer.py:54
↓ 1 callersMethoddetokenize
Detokenize tokens.
Megatron-LM/megatron/training/tokenizer/multimodal_tokenizer.py:309
↓ 1 callersMethoddiff
Get git diff from the container. Returns the patch string.
swe-rl/swe_env_client.py:97
↓ 1 callersMethoddiff
(self, lease_id: str, cwd: str = "/testbed")
swe-rl/server/swe_env_pool_server.py:156
↓ 1 callersFunctiondiff_text_file
(result: str, expect: str)
gui-rl/desktop_env/evaluators/metrics/general.py:84
↓ 1 callersFunctiondisable_forward_pre_hook
(model_chunks, param_sync=True)
Megatron-LM/tests/unit_tests/test_fp8_param.py:53
↓ 1 callersFunctiondisable_jit_fuser
Disable the JIT fuser
Megatron-LM/megatron/core/jit.py:27
↓ 1 callersFunctiondiscover_generators
Automatically discover generator modules in the generator directory. Returns a dictionary mapping task_type to module with run_rollout functi
slime/slime_plugins/rollout_buffer/buffer.py:54
↓ 1 callersMethoddispatch
( self, hidden_states: torch.Tensor, async_finish: bool = False, allocate_on_c
Megatron-LM/megatron/core/transformer/moe/token_dispatcher.py:1170
↓ 1 callersMethoddispatch
(self, topk_ids)
slime/slime/backends/fsdp_utils/models/qwen3_moe.py:62
↓ 1 callersMethoddispatcher_drop_and_pad_test
Test if the tokens are dropped and padded correctly. Since the probs of padded tokens are 0, the combined results for dispatching wit
Megatron-LM/tests/unit_tests/transformer/moe/test_token_dispatcher.py:214
↓ 1 callersFunctiondistribute_shards_to_ranks
Computes uniform distribution of workload across ranks, based on sizes. Currently, the assignment is greedy, based on: 1. Cross-parallelizati
Megatron-LM/megatron/core/dist_checkpointing/exchange_utils.py:121
↓ 1 callersFunctiondistributed_masked_whiten
Performs whitening on a tensor using global statistics from all participating GPUs. It calculates the global mean and variance across all ra
slime/slime/utils/distributed_utils.py:94
↓ 1 callersFunctiondo_setup
(odir)
Megatron-LM/tests/unit_tests/data/test_builder.py:35
↓ 1 callersFunctiondo_test_preprocess_mmdata
(temp_dir, extra_args=[])
Megatron-LM/tests/unit_tests/data/test_preprocess_mmdata.py:58
↓ 1 callersFunctiondownload_data
(ds_name)
terminal-rl/data_utils/download.py:120
↓ 1 callersFunctiondownload_job_assets
(logs: List[jet_log.JETLog], iteration: int = 0)
Megatron-LM/tests/test_utils/python_scripts/launch_jet_workload.py:158
↓ 1 callersFunctiondownload_screenshot
(vm_name)
gui-rl/desktop_env/providers/virtualbox/manager.py:255
↓ 1 callersFunctiondownload_screenshot
(ip)
gui-rl/desktop_env/providers/vmware/manager.py:248
↓ 1 callersMethoddp_rank_context
(self)
slime/slime/rollout/sglang_rollout.py:70
↓ 1 callersFunctiondrain_embedding_wgrad_compute
Helper for performing embedding wgrad GEMM's during the pipeline drain phase, pipelines the AllGather and GEMM's. Should only be used when pi
Megatron-LM/megatron/core/utils.py:1073
↓ 1 callersMethoddrain_eval_scores
(self)
openclaw-rl/openclaw_api_server.py:369
↓ 1 callersMethoddrain_eval_scores
(self)
openclaw-rl/oel/openclaw_oel_api_server.py:1568
↓ 1 callersMethoddrain_eval_scores
(self)
openclaw-opd/openclaw_opd_api_server.py:915
↓ 1 callersFunctiondrain_output_queue
Block until batch_size sample groups are collected.
openclaw-tinker/rollout.py:99
↓ 1 callersFunctiondrain_queue
Block until batch_size samples are collected from the output queue.
openclaw-fireworks/fireworks_training_loop.py:28
↓ 1 callersFunctiondummy_img
(odir_txt, odir_img)
Megatron-LM/tests/unit_tests/data/test_preprocess_mmdata.py:19
↓ 1 callersMethoddummy_step
The dummy step can be used to initialize the potential optimizer.state, which can solve the problem of checkpoint loading for an inpl
Megatron-LM/megatron/core/optimizer/cpu_offloading/hybrid_optimizer.py:453
↓ 1 callersFunctiondummy_train_step
Single dummy training step.
Megatron-LM/megatron/training/training.py:1367
↓ 1 callersMethodduplicate_requests
(self, request: InferenceRequest, n: int)
Megatron-LM/megatron/rl/inference/inference_interface.py:49
↓ 1 callersMethoddynamic_inference
Executes dynamic inference by separating decode and prefill requests and running them independently. Also runs the chunked prefill re
Megatron-LM/megatron/core/ssm/mamba_mixer.py:447
↓ 1 callersFunctiondynamic_preprocess
( image, min_num=1, max_num=6, image_size=448, use_thumbnail=False, find_closest_aspect_ratio_fn=find_
Megatron-LM/examples/multimodal/image_processing.py:88
↓ 1 callersFunctioneager_attn_bwd
Backward pass for eager attention
Megatron-LM/megatron/core/transformer/dot_product_attention_context_parallel.py:56
↓ 1 callersFunctioneager_attn_fwd
Forward pass for eager attention
Megatron-LM/megatron/core/transformer/dot_product_attention_context_parallel.py:19
↓ 1 callersMethodelapsed
This method is called from report(), or can be called directly It is called to collect all the elapsed time since last reset(). It
Megatron-LM/megatron/core/utils.py:1524
← previousnext →3,701–3,800 of 10,561, ranked by callers