MCPcopy Create free account

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

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

↓ 2 callersMethodssm_decode
Performs SSM computation for inference decode step. Args: zxBCdt: The input tensor of shape (l, b, d), which is a concat
Megatron-LM/megatron/core/ssm/mamba_mixer.py:819
↓ 2 callersMethodstart_grad_sync
Initiates grad sync (all-reduce or reduce-scatter) communication operations for all buckets in the bucket group. When ddp_co
Megatron-LM/megatron/core/distributed/param_and_grad_buffer.py:363
↓ 2 callersMethodstate
(self)
Megatron-LM/megatron/legacy/data/realm_index.py:38
↓ 2 callersMethodstate_dict_for_save_checkpoint
Save dict with state dicts of each of the models.
Megatron-LM/megatron/legacy/model/realm_model.py:91
↓ 2 callersMethodstate_dict_for_save_checkpoint
For easy load when model is combined with other heads, add an extra key.
Megatron-LM/megatron/legacy/model/t5_model.py:165
↓ 2 callersMethodstate_dict_for_save_checkpoint
Save dict with state dicts of each of the models.
Megatron-LM/megatron/legacy/model/biencoder_model.py:143
↓ 2 callersMethodstatus
(self)
gui-rl/env_pool_server.py:298
↓ 2 callersMethodstatus
(self)
swe-rl/server/swe_env_pool_server.py:190
↓ 2 callersMethodstep
(self, lease_id: str, action: Any, sleep_after_execution: float)
gui-rl/env_client.py:55
↓ 2 callersMethodstep
(self, response_text: str)
slime/examples/geo3k_vlm_multi_turn/env_geo3k.py:182
↓ 2 callersMethodstep_with_ready_grads
Step the optimizer with ready gradients, return successful.
Megatron-LM/megatron/core/optimizer/optimizer.py:943
↓ 2 callersMethodstop
(self)
slime/slime/utils/profile_utils.py:99
↓ 2 callersMethodstop_emulator
Method to stop the emulator.
gui-rl/desktop_env/providers/base.py:40
↓ 2 callersFunctionstream_acquire_context
Stream acquire context
Megatron-LM/megatron/core/pipeline_parallel/utils.py:117
↓ 2 callersFunctionstrip_query
(url)
gui-rl/desktop_env/evaluators/metrics/chrome.py:57
↓ 2 callersMethodsuspend_engines
Sends a signal to pause all inference engines.
Megatron-LM/megatron/core/inference/inference_client.py:215
↓ 2 callersMethodsuspend_resume_ctx
Context manager for of suspending and resuming the engine. This context manager records the time and memory usage when suspending and
Megatron-LM/megatron/core/inference/engines/dynamic_engine.py:480
↓ 2 callersFunctionsync_actor_critic_data
Broadcast `values` (from critic) and optionally `log_probs`/`ref_log_probs` (from actor) across PP ranks to align data dependencies. - V
slime/slime/backends/megatron_utils/data.py:749
↓ 2 callersMethodsync_all_async_calls
Check if all ranks have completed async checkpoint writing Args: is_alive (bool): if True, the current async request is not compl
Megatron-LM/megatron/core/dist_checkpointing/strategies/async_utils.py:153
↓ 2 callersMethodsynchronize_gradient_reduce
Synchronize gradient reduce-scatter operations for all model gradients.
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/megatron_fsdp.py:1076
↓ 2 callersFunctiont5_position_ids
(token_ids)
Megatron-LM/megatron/legacy/model/t5_model.py:29
↓ 2 callersFunctiont5_position_ids
Calculate position ids from token ids Args: token_ids (Tensor): input tokens Returns: Tensor: position ids
Megatron-LM/megatron/core/models/T5/t5_model.py:524
↓ 2 callersMethodtensor_need_offloading_checker
Check if the tensor needs to be offloaded.
Megatron-LM/megatron/core/pipeline_parallel/fine_grained_activation_offload.py:314
↓ 2 callersFunctiontextvqa_eval
Run TextVQA evaluation.
Megatron-LM/examples/multimodal/evaluation/evaluate_textvqa.py:38
↓ 2 callersFunctionto_float8
Convert a tensor to FP8 format.
Megatron-LM/tests/unit_tests/dist_checkpointing/test_fp8.py:20
↓ 2 callersMethodto_kitchen_qlinear
Converts to kitchen library's QLinearParams object.
Megatron-LM/megatron/core/extensions/kitchen.py:114
↓ 2 callersFunctiontoggleGif
()
swe-rl/mini-swe-agent/docs/assets/js/gif-controls.js:18
↓ 2 callersFunctiontoggle_cuda_graphs
Toggle CUDA graph-related attributes for the model and its modules. Args: set_to (str): Value to set for CUDA graph-related attribut
Megatron-LM/megatron/core/transformer/utils.py:394
↓ 2 callersMethodtoken_combine
Combines expert outputs across devices using communication. This method aggregates expert outputs from different devices via communic
Megatron-LM/megatron/core/transformer/moe/token_dispatcher.py:170
↓ 2 callersMethodtoken_dispatch
Dispatches tokens to expert devices using communication. This method performs the main communication (e.g., All-to-All) to send token
Megatron-LM/megatron/core/transformer/moe/token_dispatcher.py:112
↓ 2 callersFunctiontoken_perf
(response_lengths, non_generation_time, key="")
slime/slime/ray/rollout.py:1253
↓ 2 callersMethodtoken_to_id
Converts a single token to it's id.
Megatron-LM/megatron/core/tokenizers/text/libraries/tiktoken_tokenizer.py:170
↓ 2 callersMethodtokens_to_ids
Converts tokens to ids.
Megatron-LM/megatron/core/tokenizers/text/libraries/null_tokenizer.py:28
↓ 2 callersMethodtrack_mtp_metrics
Track the Multi-Token Prediction (MTP) metrics for logging.
Megatron-LM/megatron/core/transformer/multi_token_prediction.py:366
↓ 2 callersFunctiontrain
Run training over a rollout consisting of multiple steps. The model is switched to train mode, training hooks are configured, and ``train_one
slime/slime/backends/megatron_utils/model.py:507
↓ 2 callersMethodtrain
(self, rollout_id: int, rollout_data_ref: Box)
slime/slime/backends/megatron_utils/actor.py:615
↓ 2 callersFunctiontrain_index
Entry point for training the index. We select whether to train a new index, or validate an existing index. Args: config (RetroPrepro
Megatron-LM/megatron/core/datasets/retro/index/build.py:232
↓ 2 callersFunctiontrain_valid_test_dataloaders_provider
Build multimodal train, validation and test dataloaders.
Megatron-LM/examples/multimodal/dataloader_provider.py:95
↓ 2 callersFunctiontraining_log
Log training information such as losses, timing, ....
Megatron-LM/megatron/training/training.py:1507
↓ 2 callersFunctionunpack_sequences
Unpack sequences from a packed batch. Args: packed_batch: Packed batch Returns: List of unpacked batches
slime/slime/backends/fsdp_utils/data_packing.py:124
↓ 2 callersMethodunpause_engines
Sends a signal to unpause all inference engines.
Megatron-LM/megatron/core/inference/inference_client.py:209
↓ 2 callersFunctionupdate_loader_saver
(key)
Megatron-LM/tools/checkpoint/convert.py:134
↓ 2 callersMethodupdate_main_grads
Update the gradients in the model parameters with the main gradients from the main gradient buffer. If the model parameters are shard
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:2529
↓ 2 callersFunctionupdate_train_iters
(args)
Megatron-LM/megatron/training/training.py:920
↓ 2 callersFunctionupdate_use_dist_ckpt
(args)
Megatron-LM/megatron/training/utils.py:716
↓ 2 callersFunctionupdate_weights_from_distributed
Send metadata (Ray), broadcast tensors (NCCL rank 0 → engines).
slime/slime/backends/megatron_utils/update_weight/update_weight_from_distributed.py:287
↓ 2 callersFunctionvalidate_added_encodings
Validate added encodings. Steps: - Randomly sample subset of text dataset blocks. - Encode each block. - Compare against saved encodi
Megatron-LM/megatron/core/datasets/retro/index/validate.py:109
↓ 2 callersFunctionvalidate_integrity_and_strict_load
Validates sharding integrity and potential mismatches with the checkpoint. `validate_access_integrity` controls sharding integrity check (orthogo
Megatron-LM/megatron/core/dist_checkpointing/validation.py:128
↓ 2 callersFunctionvalidate_training_embeddings
Validate training embeddings. Steps: - Randomly sample subset of text dataset blocks. - Embed each block. - Compare against saved emb
Megatron-LM/megatron/core/datasets/retro/index/validate.py:41
↓ 2 callersFunctionvanilla_gae
( rewards: torch.Tensor, values: torch.Tensor, gamma: float, lambd: float, )
slime/slime/utils/ppo_utils.py:482
↓ 2 callersFunctionverify
Verify if the solution is correct. Args: solution_str: The solution string to verify answer: The ground truth answer stri
slime/slime/rollout/rm_hub/math_dapo_utils.py:240
↓ 2 callersFunctionvocab_parallel_cross_entropy
Performs cross entropy loss when logits are split across tensor parallel ranks Args: vocab_parallel_logits: logits split across tens
Megatron-LM/megatron/core/tensor_parallel/cross_entropy.py:219
↓ 2 callersMethodvocab_range_from_per_partition_vocab_size
Vocab range from per partition vocab size.
Megatron-LM/megatron/core/tensor_parallel/utils.py:105
↓ 2 callersMethodwait_and_update_bucket_weights
(self, bucket)
slime/slime/backends/fsdp_utils/update_weight_utils.py:79
↓ 2 callersFunctionweighted_bias_quick_geglu_impl
Token-wise-weighted bias quick_geglu fusion. input: [num_selected_experts * seq_len, hidden_size * 2] bias: None weights:
Megatron-LM/megatron/core/fusions/fused_bias_geglu.py:410
↓ 2 callersFunctionweighted_squared_relu_impl
Token-wise weighted Squared-ReLU fusion with optional FP8 storage. Args: input (torch.Tensor): Input tensor of shape ``(B, *, hidden_size
Megatron-LM/megatron/core/fusions/fused_weighted_squared_relu.py:91
↓ 2 callersFunctionwgrad_compute
(all_gathered_input, grad_output, weight)
Megatron-LM/megatron/core/utils.py:1108
↓ 2 callersFunctionwindow_partition
Args: x: (B, H, W, C) window_size (int): window size Returns: windows: (num_windows*B, window_size, window_size, C)
Megatron-LM/megatron/legacy/model/vision/esvit_swin_backbone.py:45
↓ 2 callersFunctionwindow_partition
Args: x: (B, H, W, C) window_size (int): window size Returns: windows: (num_windows*B, window_size, window_size, C)
Megatron-LM/megatron/legacy/model/vision/swin_backbone.py:39
↓ 2 callersMethodworker_url
(self, worker_idx: int)
terminal-rl/router_server.py:82
↓ 2 callersMethodzero_parameters
Zero out all parameters in embedding.
Megatron-LM/megatron/legacy/model/language_model.py:192
↓ 1 callersMethod__call__
Calls the forward pass of the cudagraphed module. Args: megatron_module (torch.nn.module): The megatron module to be graphed and
Megatron-LM/megatron/core/transformer/cuda_graphs.py:1250
↓ 1 callersMethod__create_chunk_list__
Delegates creating chunk items to local shards. Returns: List[ChunkStorageMetadata]: list of DCP ChunkStorageMetadata metadata ob
Megatron-LM/megatron/core/dist_checkpointing/strategies/checkpointable.py:155
↓ 1 callersFunction__download_and_unzip_vm
()
gui-rl/desktop_env/providers/virtualbox/manager.py:50
↓ 1 callersFunction__download_and_unzip_vm
()
gui-rl/desktop_env/providers/vmware/manager.py:122
↓ 1 callersMethod__enter__
(self)
Megatron-LM/megatron/core/nccl_allocator.py:190
↓ 1 callersMethod__enter__
(self)
slime/slime/rollout/rm_hub/math_dapo_utils.py:74
↓ 1 callersMethod__exit__
(self, *args)
Megatron-LM/megatron/core/nccl_allocator.py:207
↓ 1 callersMethod__exit__
(self, *args)
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:215
↓ 1 callersMethod__exit__
(self, type, value, tb)
Megatron-LM/megatron/training/dist_signal_handler.py:72
↓ 1 callersMethod__exit__
(self, type, value, traceback)
slime/slime/rollout/rm_hub/math_dapo_utils.py:78
↓ 1 callersMethod__getattribute__
Intercepts calls like a.hello_world()
Megatron-LM/megatron/core/utils.py:281
↓ 1 callersMethod__init__
( self, *, max_tasks: int, max_runs_per_task: int, run_idle_ttl: int,
terminal-rl/remote/pool_server.py:81
↓ 1 callersMethod__init__
( self, *, model_type: str, sglang_client: SGLangClient, non_think_mod
terminal-rl/agent/camel_agent.py:318
↓ 1 callersMethod__init__
This "meta"-model randomly selects one of the models at every call
swe-rl/mini-swe-agent/src/minisweagent/models/extra/roulette.py:17
↓ 1 callersMethod__init__
(self, ict_head_size, num_tokentypes=1, parallel_output=Tru
Megatron-LM/megatron/legacy/model/realm_model.py:41
↓ 1 callersMethod__init__
(self, config, num_tokentypes=0, parallel_output=True,
Megatron-LM/megatron/legacy/model/t5_model.py:67
↓ 1 callersMethod__init__
(self, num_tokentypes=1, parallel_output=True, only_query_m
Megatron-LM/megatron/legacy/model/biencoder_model.py:68
↓ 1 callersMethod__init__
(self, config, num_tokentypes=2, add_binary_head=True,
Megatron-LM/megatron/legacy/model/bert_model.py:130
↓ 1 callersMethod__init__
(self, config, num_classes, finetune=False, pre_process=True, post_process=True)
Megatron-LM/megatron/legacy/model/vision/classification.py:16
↓ 1 callersMethod__init__
(self, config, hidden_size, num_classes)
Megatron-LM/megatron/legacy/model/vision/vit_backbone.py:33
↓ 1 callersMethod__init__
( self, root: str, transform: Optional[Callable] = None, targe
Megatron-LM/megatron/legacy/data/image_folder.py:285
↓ 1 callersMethod__init__
(self, m, n)
Megatron-LM/megatron/legacy/mpu/tests/test_layers.py:164
↓ 1 callersMethod__init__
Initialize Timer. Args: name (str): Name of the timer.
Megatron-LM/megatron/core/timers.py:121
↓ 1 callersMethod__init__
(self, *args, **kwargs)
Megatron-LM/megatron/core/utils.py:277
↓ 1 callersMethod__init__
( self, dim: Optional[int] = None, denominator_offset: Union[torch.Tensor, float] = 1.0 )
Megatron-LM/megatron/core/fusions/fused_softmax.py:161
↓ 1 callersMethod__init__
( self, input_size: int, output_size: int, *, config: ModelParallelCon
Megatron-LM/megatron/core/post_training/modelopt/layers.py:101
↓ 1 callersMethod__init__
( self, input_size: int, output_size: int, *, config: ModelParallelCon
Megatron-LM/megatron/core/tensor_parallel/inference_layers.py:105
↓ 1 callersMethod__init__
(self)
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/utils.py:662
↓ 1 callersMethod__init__
( self, block_count_total, max_kv_block_count, max_requests, block_size_tokens, max_seqlen )
Megatron-LM/megatron/core/inference/contexts/attention_context/mha_metadata.py:14
↓ 1 callersMethod__init__
( self, strategy: SaveShardedStrategy, parallelization_group: Optional[torch.distribut
Megatron-LM/megatron/core/dist_checkpointing/strategies/fully_parallel.py:73
↓ 1 callersMethod__init__
(self, backend: str, version: int)
Megatron-LM/megatron/core/dist_checkpointing/strategies/zarr.py:78
↓ 1 callersMethod__init__
( self, indexed_dataset: IndexedDataset, dataset_path: Optional[str], indexed_
Megatron-LM/megatron/core/datasets/gpt_dataset.py:99
↓ 1 callersMethod__init__
(self, idx_path: str, dtype: Type[numpy.number])
Megatron-LM/megatron/core/datasets/indexed_dataset.py:130
↓ 1 callersMethod__init__
(self, block_paths: List[str], block_size: int)
Megatron-LM/megatron/core/datasets/retro/utils.py:357
↓ 1 callersMethod__init__
( self, config: TransformerConfig, encoder_config: TransformerConfig, transfor
Megatron-LM/megatron/core/models/T5/t5_model.py:132
↓ 1 callersMethod__init__
( self, kv_channels: int, rotary_percent: float, rotary_interleaved: bool = Fa
Megatron-LM/megatron/core/models/common/embeddings/rotary_pos_embedding.py:58
↓ 1 callersMethod__init__
(self, config)
Megatron-LM/megatron/core/models/huggingface/module.py:20
↓ 1 callersMethod__init__
(self, config: RetroConfig, submodules: Type, **kwargs: dict)
Megatron-LM/megatron/core/models/retro/encoder_attention.py:201
↓ 1 callersMethod__init__
( self, config: RetroConfig, submodules: CrossAttentionSubmodules, layer_numbe
Megatron-LM/megatron/core/models/retro/decoder_attention.py:58
← previousnext →2,401–2,500 of 10,561, ranked by callers