MCPcopy Create free account

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

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

↓ 1 callersFunctionprepare
()
slime/tests/test_qwen3_vl_4B_fsdp.py:11
↓ 1 callersFunctionprepare
()
slime/examples/true_on_policy/run_simple.py:15
↓ 1 callersFunctionprepare
()
slime/examples/geo3k_vlm_multi_turn/run_geo3k_vlm_multi_turn.py:32
↓ 1 callersFunctionprepare
()
slime/examples/true_on_policy_vlm/run_simple.py:13
↓ 1 callersFunctionprepare_data_for_update
Extract data for the update from raw rollouts. Args: model: Current policy as the zero-eth element. ref_state_dict: Reference pol
Megatron-LM/megatron/rl/rl_utils.py:1158
↓ 1 callersMethodprepare_decentralized_global_plan
Instead of assigning indices by plan order, uses PyT rank (same outcome). Args: local_plan (SavePlan): local plan to turn to a gl
Megatron-LM/megatron/core/dist_checkpointing/strategies/filesystem_async.py:460
↓ 1 callersMethodprepare_for_absorption
Prepare the model for absorption optimization in MLA (Multi-Latent Attention). This method sets up the necessary components for the absorptio
Megatron-LM/megatron/core/transformer/multi_latent_attention.py:920
↓ 1 callersMethodprepare_grads
Pre-processing gradients before the optimizer step, returns whether inf/nan is found.
Megatron-LM/megatron/core/optimizer/optimizer.py:524
↓ 1 callersMethodprepare_grads
Pre-processing gradients before the optimizer step, returns whether inf/nan is found.
Megatron-LM/megatron/core/optimizer/optimizer.py:1233
↓ 1 callersFunctionprepare_homework_files
Write each problem to workspace/homework/i.txt, return count written.
openclaw-test/student_chat.py:95
↓ 1 callersMethodprepare_local_plan
Prepare the local plan for the checkpointing process.
Megatron-LM/megatron/core/dist_checkpointing/strategies/filesystem_async.py:500
↓ 1 callersFunctionprepare_model_for_fp8_inference
Prepare a model for FP8 inference by wrapping TE linear layers with padding support. FP8 TE Gemms have specific shape requirements. This func
Megatron-LM/megatron/core/fp8_utils.py:751
↓ 1 callersFunctionprepare_packed_trajectories
Prepare trajectories for sequence packing mode with distributed processing. Distributes trajectory preparation across ranks, then gathers results
Megatron-LM/megatron/rl/rl_utils.py:1051
↓ 1 callersMethodprepare_write_data
First stage of async saving. Copy data to CPU and plan the local saving. Args: plan (SavePlan): save plan generated by t
Megatron-LM/megatron/core/dist_checkpointing/strategies/filesystem_async.py:104
↓ 1 callersMethodpreprocess
Preprocesses the token routing map for All-to-All communication and token permutation. This method computes the number of tokens ass
Megatron-LM/megatron/core/transformer/moe/token_dispatcher.py:472
↓ 1 callersFunctionpreprocess_dcp_and_te_extra_state
(state_dict)
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/fully_shard.py:215
↓ 1 callersFunctionpreprocess_image
Example image preprocessing. Resizes input image to target size. Args: target_h (int): Target height in pixels. target_w (int): T
Megatron-LM/tools/run_vlm_text_generation.py:49
↓ 1 callersMethodpretty_repr
Pretty representation of the custom layout, showing the layers held by each stage. Example: VPP rank 0
Megatron-LM/megatron/core/transformer/pipeline_parallel_layer_layout.py:207
↓ 1 callersFunctionprint_diff_in_state_dicts
Print the differences between two state dicts: metadata state dict and load state dict. This function compares the keys and shapes of the ten
Megatron-LM/megatron/core/transformer/fsdp_dtensor_checkpoint.py:370
↓ 1 callersFunctionprint_memory_usage
Print memory usage.
Megatron-LM/tools/checkpoint/utils.py:43
↓ 1 callersMethodprint_report
Print complete list of instances and their exit statuses.
swe-rl/mini-swe-agent/src/minisweagent/run/extra/utils/batch_progress.py:161
↓ 1 callersMethodprint_usage
Print usage.
Megatron-LM/tools/retro/cli/cli.py:210
↓ 1 callersMethodprocess_image_tag
(self, q)
Megatron-LM/examples/multimodal/evaluation/evaluation_datasets.py:258
↓ 1 callersFunctionprocess_mrs
Process all MRs from given milestones.
Megatron-LM/tests/test_utils/python_scripts/auto_reminder.py:272
↓ 1 callersMethodprocess_samples_from_single_path
(filename)
Megatron-LM/megatron/legacy/data/orqa_wiki_dataset.py:164
↓ 1 callersFunctionprocess_single_sample
(data)
Megatron-LM/examples/multimodal/evaluation/mmmu_utils.py:73
↓ 1 callersFunctionprocess_tool_calls
( text: str, tools: list[Any], tool_call_parser: str | None, finish_reason: str, use_respo
terminal-rl/inference_client.py:28
↓ 1 callersMethodproject_embeddings
Project image embeddings using input or output projections. Args: embeddings: List of image embeddings to project is_
Megatron-LM/megatron/core/models/mimo/submodules/vision.py:135
↓ 1 callersMethodproject_embeddings
Project embeddings to the language model dimension space.
Megatron-LM/megatron/core/models/mimo/submodules/audio.py:108
↓ 1 callersFunctionpropagate_no_split_modules
Ensure ``_no_split_modules`` is visible through the PEFT wrapper. PEFT wraps the original model as ``model.base_model.model``. FSDP's ``appl
slime/slime/backends/fsdp_utils/lora_utils.py:53
↓ 1 callersMethodprovider_bridge
(self, hf_pretrained)
slime/slime_plugins/megatron_bridge/qwen3_5.py:35
↓ 1 callersFunctionpull_docker_image
(image: str, timeout: float = 1200.0)
terminal-rl/remote/docker_compose_utils.py:160
↓ 1 callersMethodpurge_record_files
Clear all record JSONL files. Called when training starts.
openclaw-rl/openclaw_api_server.py:380
↓ 1 callersMethodpurge_record_files
(self)
openclaw-rl/oel/openclaw_oel_api_server.py:1578
↓ 1 callersMethodpurge_record_files
(self)
openclaw-opd/openclaw_opd_api_server.py:925
↓ 1 callersMethodput
Handle generation request.
Megatron-LM/megatron/core/inference/text_generation_server/text_generation_server.py:34
↓ 1 callersMethodqa_template
(self, data)
Megatron-LM/examples/multimodal/evaluation/evaluation_datasets.py:1262
↓ 1 callersFunctionquant_fp8
(weight, strategy, block_size=None)
slime/tools/convert_hf_to_fp8.py:89
↓ 1 callersFunctionquantile
(total_value, n_quantiles, data)
slime/slime/backends/megatron_utils/data.py:622
↓ 1 callersFunctionquantize
( x, scale, zero_point, dtype=torch.int8, )
slime/tools/convert_hf_to_int4_direct.py:90
↓ 1 callersFunctionquantize
( x, scale, zero_point, dtype=torch.int8, )
slime/slime/backends/megatron_utils/megatron_to_hf/processors/quantizer_compressed_tensors.py:195
↓ 1 callersFunctionquantize_param_shard
Cast shard fp32 main params to fp8 model params.
Megatron-LM/megatron/core/fp8_utils.py:484
↓ 1 callersFunctionquantize_params
(args, megatron_name, converted_named_params, quantization_config)
slime/slime/backends/megatron_utils/megatron_to_hf/processors/__init__.py:8
↓ 1 callersFunctionquantize_params_compressed_tensors
(converted_named_params, quantization_config)
slime/slime/backends/megatron_utils/megatron_to_hf/processors/quantizer_compressed_tensors.py:266
↓ 1 callersFunctionquantize_params_fp8
(args, megatron_name, converted_named_params, quantization_config)
slime/slime/backends/megatron_utils/megatron_to_hf/processors/quantizer_fp8.py:10
↓ 1 callersMethodquery
(self, messages: list[dict], **kwargs)
swe-rl/mini-swe-agent/src/minisweagent/models/anthropic.py:15
↓ 1 callersMethodquery
(self)
swe-rl/mini-swe-agent/src/minisweagent/agents/interactive_textual.py:52
↓ 1 callersMethodquery
Query the model and return the response.
swe-rl/mini-swe-agent/src/minisweagent/agents/default.py:92
↓ 1 callersFunctionquery_block_neighbors
Query neighbors of a dataset block (i.e., range). Args: config (RetroPreprocessingConfig): Retro preprocessing config. db_dataset
Megatron-LM/megatron/core/datasets/retro/query/query.py:261
↓ 1 callersFunctionquery_dataset_neighbors
Query neighbors of each chunk within a dataset. Args: config (RetroPreprocessingConfig): Retro preprocessing config. db_dataset (
Megatron-LM/megatron/core/datasets/retro/query/query.py:318
↓ 1 callersFunctionquery_embedding_block
Query a block of embeddings. The block is broken into smaller sub-blocks, for easier tracking of progress. Both the raw neighbor IDs and the
Megatron-LM/megatron/core/datasets/retro/query/query.py:189
↓ 1 callersFunctionquery_embeddings
Query neighbors of a block of embeddings. Querying includes: - Query index for neighbor chunk IDs. - Filter chunk IDs that have the s
Megatron-LM/megatron/core/datasets/retro/query/query.py:116
↓ 1 callersFunctionquery_neighbors
Query pretraining datasets (train & valid). Args: config (RetroPreprocessingConfig): Retro preprocessing config.
Megatron-LM/megatron/core/datasets/retro/query/query.py:405
↓ 1 callersFunctionquick_gelu
Sigmoid approximation of gelu
Megatron-LM/megatron/core/fusions/fused_bias_geglu.py:185
↓ 1 callersFunctionrandbool
(shape, **kwargs)
Megatron-LM/tests/unit_tests/transformer/test_attention.py:828
↓ 1 callersFunctionrdtablebench_eval
Run RD-TableBench evaluation.
Megatron-LM/examples/multimodal/evaluation/evaluate_rd_tablebench.py:42
↓ 1 callersFunctionread_audio
Process audio file and return tensor.
Megatron-LM/examples/mimo/avlm_inference.py:44
↓ 1 callersFunctionread_image
Process image file and return tensor.
Megatron-LM/examples/mimo/avlm_inference.py:61
↓ 1 callersFunctionread_json
(path)
Megatron-LM/tools/checkpoint/loader_llama_mistral.py:65
↓ 1 callersFunctionread_prediction
(prediction_file)
Megatron-LM/tools/retro/text_generation/evaluate.py:90
↓ 1 callersFunctionread_x11_image
Pure Python X11 (XWD) format reader that converts to PIL Image. No external dependencies required. Args: filepath: Path to t
gui-rl/desktop_env/evaluators/metrics/docs.py:28
↓ 1 callersMethodreceive_checkpoint_metadata
Receive and populate model metadata.
Megatron-LM/tools/checkpoint/saver_base.py:265
↓ 1 callersMethodreceive_model
Creates model scheme and receives model over the queue
Megatron-LM/tools/checkpoint/saver_base.py:614
↓ 1 callersMethodreceive_model
(self)
Megatron-LM/tools/checkpoint/saver_hf_llava.py:327
↓ 1 callersMethodreceive_vision_backbone
(self, schema)
Megatron-LM/tools/checkpoint/saver_hf_llava.py:69
↓ 1 callersMethodreceive_vision_backbone
(self, schema)
Megatron-LM/tools/checkpoint/saver_llava.py:184
↓ 1 callersMethodreceive_vision_projection
(self)
Megatron-LM/tools/checkpoint/saver_hf_llava.py:176
↓ 1 callersMethodreceive_vision_projection
(self)
Megatron-LM/tools/checkpoint/saver_llava.py:321
↓ 1 callersFunctionreconfigure_num_microbatches_calculator
Reconfigure number of microbatches calculator. Supporting backward compatibility. Args: rank (int): Rank of the GPU, only ran
Megatron-LM/megatron/core/num_microbatches_calculator.py:107
↓ 1 callersMethodrecord_bwd_graph
Record a bwd graph to 'cudagraph_record
Megatron-LM/megatron/core/transformer/cuda_graphs.py:199
↓ 1 callersMethodrecord_graph_capture
Records the data needed to create this runner's forward cudagraph. The first pass records a graph and appends the runner to _CudagraphGlobalRe
Megatron-LM/megatron/core/transformer/cuda_graphs.py:858
↓ 1 callersMethodrecord_model_turn
( self, turn_idx: int, *, assistant_text: str, tool_calls: Optional[Li
terminal-rl/rollout_agent.py:33
↓ 1 callersMethodrecover_lm_qkv_bias
(self, qkv_bias)
Megatron-LM/tools/checkpoint/saver_hf_llava.py:225
↓ 1 callersMethodrecover_lm_qkv_weight
(self, qkv_weight)
Megatron-LM/tools/checkpoint/saver_hf_llava.py:188
↓ 1 callersFunctionrecursively_lint_files
Recursively lint all python files in chosen subdirectories of megatron-lm
Megatron-LM/tools/linter.py:7
↓ 1 callersMethodrecv_backward
Receive backward gradient tensor. Note: Gradient receivers are activation 'SENDERS' Args: tensor_shape: Expected gradien
Megatron-LM/megatron/core/pipeline_parallel/bridge_communicator.py:451
↓ 1 callersFunctionreduce_aux_losses_tracker_across_ranks
Collect and reduce the auxiliary losses across ranks.
Megatron-LM/megatron/core/transformer/moe/moe_utils.py:763
↓ 1 callersMethodreduce_loss_in_tracker
Collect and reduce the mtp losses across ranks.
Megatron-LM/megatron/core/transformer/multi_token_prediction.py:352
↓ 1 callersMethodreduce_loss_in_tracker
Collect and reduce the indexer losses across ranks.
Megatron-LM/megatron/core/transformer/experimental_attention_variant/dsa.py:92
↓ 1 callersMethodreduce_scatter_gradients
Reduce scatter the gradients. Args: async_op (bool, optional): Whether to do the all-reduce asynchronously. Defaul
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:2807
↓ 1 callersFunctionreduce_scatter_with_fp32_accumulation
Reduce-scatter with FP32 accumulation. Collects input_tensor in lower precision using an all-to-all, then locally accumulates in FP32 precisi
Megatron-LM/megatron/core/distributed/reduce_scatter_with_fp32_accumulation.py:42
↓ 1 callersFunctionreformat_prompt_retro
(query, neighbours, dataset_name, ft_neighbours, \ max_output_len, tokenizer, max_se
Megatron-LM/tools/retro/sft/dataset_conv.py:267
↓ 1 callersFunctionreformat_prompt_short
(query, neighbours, dataset_name, ft_neighbours, \ max_output_len, tokenizer, max_se
Megatron-LM/tools/retro/sft/dataset_conv.py:399
↓ 1 callersFunctionreformulate_single_nd_flattened_tensor
Reformulates shapes of a single N-D flattened ShardedTensor. We need to define a pair of transformations: - turn N-D ShardedTensor with origi
Megatron-LM/megatron/core/dist_checkpointing/strategies/resharding.py:185
↓ 1 callersFunctionregister_default_common_strategies
Register default common strategies.
Megatron-LM/megatron/core/dist_checkpointing/strategies/common.py:29
↓ 1 callersFunctionregister_default_tensorstore_strategies
Register default strategies leveraging tensorstore.
Megatron-LM/megatron/core/dist_checkpointing/strategies/tensorstore.py:33
↓ 1 callersFunctionregister_default_torch_strategies
Register default strategies related to PyT Distributed backend.
Megatron-LM/megatron/core/dist_checkpointing/strategies/torch.py:94
↓ 1 callersFunctionregister_default_zarr_strategies
Register default strategies related to Zarr backend.
Megatron-LM/megatron/core/dist_checkpointing/strategies/zarr.py:68
↓ 1 callersFunctionregister_pipeline_terminator
(pipeline: jetclient.JETPipeline)
Megatron-LM/tests/test_utils/python_scripts/launch_jet_workload.py:34
↓ 1 callersFunctionregister_routing_replay
(module)
slime/slime/utils/routing_replay.py:85
↓ 1 callersFunctionregister_safe_globals
Register megatron-core safe classes with torch serialization.
Megatron-LM/megatron/core/safe_globals.py:39
↓ 1 callersMethodregister_tool
Register a new tool in the registry
toolcall-rl/tool_sandbox.py:329
↓ 1 callersFunctionregroup
Rebuild the nested dict from [(key_path, tensor), …].
Megatron-LM/examples/mimo/utils/data_helpers.py:32
↓ 1 callersFunctionrelative_grad_diff
Relative difference ||g_hat - g_ref||_F / ||g_ref||_F. Accepts a single tensor or an iterable of shards for each argument.
Megatron-LM/tests/functional_tests/python_test_utils/test_optimizer_grads_match.py:41
↓ 1 callersMethodrelease
(self)
Megatron-LM/megatron/training/dist_signal_handler.py:75
↓ 1 callersFunctionreload_mergeable_ranks
Reload the tokenizer JSON file and convert it to Tiktoken format. Args: path (str): path to the tokenizer. max_vocab (Option
Megatron-LM/megatron/core/tokenizers/text/libraries/tiktoken_tokenizer.py:26
↓ 1 callersFunctionreload_mergeable_ranks
Reload our tokenizer JSON file and convert it to Tiktoken format.
Megatron-LM/megatron/training/tokenizer/tokenizer.py:636
↓ 1 callersMethodreload_process_groups
()
slime/slime/utils/reloadable_process_group.py:149
↓ 1 callersFunctionremote_rm
(args, sample: Sample)
slime/slime/rollout/rm_hub/__init__.py:17
↓ 1 callersFunctionremove_articles
(text)
Megatron-LM/tools/retro/text_generation/evaluate.py:18
← previousnext →4,801–4,900 of 10,561, ranked by callers