MCPcopy Create free account

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

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

Functiontopk_distillation_loss_function
Compute top-K logits-based distillation loss with tail trick. Computes reverse KL divergence D_KL(student || teacher) over the teacher's top-
openclaw-opd/topk_distillation_loss.py:26
Methodtorch_storage
()
Megatron-LM/tests/unit_tests/fusions/test_fused_linear_cross_entropy.py:484
Methodtorch_storage
()
Megatron-LM/tests/unit_tests/fusions/test_fused_linear_cross_entropy.py:908
Methodtorch_storage
()
Megatron-LM/tests/unit_tests/fusions/test_fused_linear_cross_entropy.py:1419
Functiontrace_async_exceptions
Decorator to be applied to every coroutine that runs in a separate task. This is needed because asyncio tasks do not propagate exceptions. Co
Megatron-LM/megatron/core/utils.py:2288
Functiontrack_app_tag
Track app_tag and app_tag ID Args: batch_size (int): current batch size world_size (int): the number of processes of current job
Megatron-LM/megatron/training/one_logger_utils.py:444
Functiontrack_config_flags
Track flags about train/validation/test enablement Args: train_iters (int): target train iteration number skip_train (bool): flag
Megatron-LM/megatron/training/one_logger_utils.py:300
Functiontrack_e2e_metrics
Track E2E application metrics with one-logger NOTE: the function should be called after barrier call. Args: log_throughput (bool, op
Megatron-LM/megatron/training/one_logger_utils.py:209
Methodtrack_forward
(x)
Megatron-LM/tests/unit_tests/test_fp8_utils.py:94
Methodtracked_metadata
Obtain an ordered list of all request metadata to be tracked by the context. This consists of metadata that is used to inform text generation
Megatron-LM/megatron/core/inference/inference_request.py:299
Methodtrain
(self, rollout_id, rollout_data_ref)
slime/slime/ray/train_actor.py:112
Functiontrain_iters
Simple fixture returning number of train iters.
Megatron-LM/tests/functional_tests/python_test_utils/conftest.py:78
Methodtrain_step
(model, optimizer, inputs)
Megatron-LM/tests/unit_tests/distributed/test_mcore_fully_sharded_data_parallel.py:191
Functiontrain_valid_test_datasets_provider
Build train, valid, and test datasets.
Megatron-LM/pretrain_bert.py:145
Functiontrain_valid_test_datasets_provider
Build the train test and validation datasets. Args: train_val_test_num_samples : A list containing the number of samples in t
Megatron-LM/pretrain_t5.py:201
Functiontrain_valid_test_datasets_provider
Build the train test and validation datasets. Args: train_val_test_num_samples : A list containing the number of samples in train test an
Megatron-LM/pretrain_gpt.py:252
Functiontrain_valid_test_datasets_provider
Build the train test and validation datasets. Args: train_val_test_num_samples : A list containing the number of samples in train, valida
Megatron-LM/pretrain_vlm.py:204
Functiontrain_valid_test_datasets_provider
Build train, valid and test datasets.
Megatron-LM/pretrain_ict.py:139
Functiontrain_valid_test_datasets_provider
Build the train test and validation datasets. Args: train_val_test_num_samples : A list containing the number of samples in train test an
Megatron-LM/pretrain_mamba.py:182
Functiontrain_valid_test_datasets_provider
For GRPO, use lightweight minimal datasets instead of heavyweight mocks.
Megatron-LM/train_rl.py:315
Functiontrain_valid_test_datasets_provider
Build train, valid, and test datasets.
Megatron-LM/pretrain_retro.py:184
Functiontrain_valid_test_datasets_provider
Build train, valid, and test datasets.
Megatron-LM/pretrain_vision_dino.py:80
Functiontrain_valid_test_datasets_provider
Build train, valid, and test datasets.
Megatron-LM/pretrain_vision_inpaint.py:116
Functiontrain_valid_test_datasets_provider
Build train, valid, and test datasets.
Megatron-LM/pretrain_vision_classify.py:81
Functiontrain_valid_test_datasets_provider
Build train, valid, and test datasets.
Megatron-LM/tools/retro/sft/sft_retro.py:191
Functiontrain_valid_test_datasets_provider
Dataset provider for MIMO model training. Args: *provider_args: Additional arguments for the dataset provider **provider_kwargs:
Megatron-LM/examples/mimo/train.py:177
Functiontrain_valid_test_datasets_provider
Provide datasets for training, validation, and testing.
Megatron-LM/examples/mimo/data/mock.py:240
Functiontrain_valid_test_datasets_provider
Build train, valid, and test datasets.
Megatron-LM/examples/academic_paper_scripts/detxoify_lm/finetune_gpt.py:100
Functiontrain_valid_test_sft_datasets_provider
Build the train test and validation datasets. Args: train_val_test_num_samples : A list containing the number of samples in t
Megatron-LM/examples/post_training/modelopt/finetune.py:339
Functiontransform
(example)
toolcall-rl/rl_data_preprocess.py:8
Functiontransform
(x)
Megatron-LM/examples/multimodal/image_processing.py:159
Methodtransform_object
Make no transformations - bytes objects are already serialized.
Megatron-LM/megatron/core/dist_checkpointing/strategies/torch.py:565
Functiontransform_object_override
(write_item, obj)
Megatron-LM/tools/checkpoint/checkpoint_inspector.py:293
Methodtrim_overload
Iteratively pop the most‑recent sequence from the *most‑loaded group* whenever doing so reduces the global slack.
Megatron-LM/megatron/core/pipeline_parallel/hybrid_cp_schedule.py:294
Functiontruncate_segments
Truncates a pair of sequences to a maximum sequence length.
Megatron-LM/megatron/legacy/data/dataset_utils.py:116
Functiontuple_type
Convert a string to a tuple of integers. Examples: "1,2,3" -> (1, 2, 3) "(1,2,3)" -> (1, 2, 3)
Megatron-LM/megatron/training/arguments.py:362
Functiontwod_interpolate_position_embeddings_hook
( state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error
Megatron-LM/megatron/legacy/model/vision/vit_backbone.py:57
Methodtype_of_test_result
(self)
Megatron-LM/tests/functional_tests/python_test_utils/common.py:44
Methodtype_of_test_result
(self)
Megatron-LM/tests/functional_tests/python_test_utils/common.py:61
Methodunique_identifiers
Returns a dictionary of unique identifiers.
Megatron-LM/megatron/core/tokenizers/text/text_tokenizer.py:160
Methodunique_identifiers
Property required for use with megatron-core datasets.
Megatron-LM/megatron/core/tokenizers/text/libraries/null_tokenizer.py:45
Methodunique_key
returns a unique key for this object
Megatron-LM/megatron/core/dist_checkpointing/mapping.py:576
Methodunk
Returns id of of unknown token.
Megatron-LM/megatron/core/tokenizers/text/text_tokenizer.py:206
Methodunk
(self)
Megatron-LM/megatron/training/tokenizer/tokenizer.py:761
Methodunk_id
Returns id of of unknown token. Need for NeMo.
Megatron-LM/megatron/core/tokenizers/text/text_tokenizer.py:211
Methodunk_id
Get the unknown ID.
Megatron-LM/megatron/core/tokenizers/text/libraries/bytelevel_tokenizer.py:160
Methodunk_id
Returns id of unknown tokens.
Megatron-LM/megatron/core/tokenizers/text/libraries/huggingface_tokenizer.py:319
Methodunk_id
Returns id of unknown tokens.
Megatron-LM/megatron/core/tokenizers/text/libraries/tiktoken_tokenizer.py:256
Functionunlink_data
(x)
Megatron-LM/megatron/core/dist_checkpointing/strategies/resharding.py:158
Methodunmap_weight_tensor
(m)
Megatron-LM/megatron/core/distributed/distributed_data_parallel.py:310
Methodunwrap
Returns the wrapped tensor while deleting the internal reference. Can only be called once.
Megatron-LM/megatron/core/utils.py:656
Methodupdate
(self, consumed_samples, consistency_check, verbose=False)
Megatron-LM/megatron/core/num_microbatches_calculator.py:356
Methodupdate
Args: request_query_lengths: (>real_batch_size,) request_kv_length_offsets: (>real_batch_size,) request_t
Megatron-LM/megatron/core/inference/contexts/attention_context/mha_metadata.py:146
Methodupdate
Args: request_query_lengths: (>real_batch_size,) request_kv_length_offsets: (>real_batch_size,) request_t
Megatron-LM/megatron/core/inference/contexts/attention_context/mha_metadata.py:179
Methodupdate
Construct the metadata from request states.
Megatron-LM/megatron/core/inference/contexts/attention_context/metadata_base.py:20
Methodupdate
(self, found_inf: bool)
Megatron-LM/megatron/core/optimizer/grad_scaler.py:26
Methodupdate
(self, found_inf: bool)
Megatron-LM/megatron/core/optimizer/grad_scaler.py:43
Methodupdate_bucket_weights
(self, named_tensors, weight_version=None)
slime/slime/backends/fsdp_utils/update_weight_utils.py:124
Methodupdate_bucket_weights
Send names/dtypes/shapes metadata to engines, then broadcast tensors. Ensures tensors are contiguous; when `world_size == 1`, converts DTenso
slime/slime/backends/fsdp_utils/update_weight_utils.py:227
Methodupdate_index
Delete existing index and create a new
Megatron-LM/megatron/legacy/data/realm_index.py:178
Methodupdate_sampling_client
(self, client)
openclaw-tinker/api_server.py:433
Functionupdate_sequence_packing_metrics
Update bin tracking for sequence packing mode.
Megatron-LM/megatron/rl/rl_utils.py:2360
Methodupdate_weights
Broadcast weights from rank 0 to all other ranks.
slime/slime/ray/actor_group.py:123
Methodupdate_weights
(self)
slime/slime/backends/megatron_utils/actor.py:1442
Methodupdate_weights
version++, flush caches, process buckets. Progress on rank 0.
slime/slime/backends/megatron_utils/update_weight/update_weight_from_tensor.py:107
Methodupdate_weights
Pause → flush → non-expert (TP) → expert (EP) → continue. Progress on PP source.
slime/slime/backends/megatron_utils/update_weight/update_weight_from_distributed.py:74
Methodupdate_weights
Synchronize actor weights to rollout engines. Handles both colocated and distributed update modes. In offload mode, wakes up paramete
slime/slime/backends/fsdp_utils/actor.py:762
Methodupdate_weights
(self)
slime/slime/backends/fsdp_utils/update_weight_utils.py:47
Methodupdate_weights_from_distributed
( self, names, dtypes, shapes, group_name, flush_cache=False, weight_version: str | None = None )
slime/slime/backends/sglang_utils/sglang_engine.py:399
Methodupdate_weights_from_tensor
Update model weights from tensor data. The HTTP server will only post meta data, and the real weights will be copied directly from GPUs.
slime/slime/backends/sglang_utils/sglang_engine.py:272
Functionupload_file
()
gui-rl/desktop_env/server/main.py:1156
Methodupload_file
Upload the file to S3
Megatron-LM/megatron/core/datasets/object_storage_utils.py:53
Methodupload_file
(self, Filename: str, Bucket: str, Key: str)
Megatron-LM/tests/unit_tests/data/test_bin_reader.py:78
Functionvalidate
Validation method for validating encoding blocks. Args: f (h5py.File): File with block of encodings.
Megatron-LM/megatron/core/datasets/retro/index/validate.py:129
Methodvalidate
Validation method for validating loaded encodings. Args: f (h5py.File): File that contains encodings.
Megatron-LM/megatron/core/datasets/retro/index/indexes/faiss_par_add.py:129
Functionvalidate_case_matches_checkpoint
Checks whether the casing config is consistent with the checkpoint name.
Megatron-LM/megatron/training/tokenizer/bert_tokenization.py:28
Methodvalidate_checkpoint_id
Validate the checkpoint_id that will be used to save the checkpoint. This method is available in PyTorch 2.3 and above.
Megatron-LM/megatron/core/dist_checkpointing/strategies/filesystem_async.py:521
Functionvalidate_index
Validate index. Validating index involves sequentially running stages above: - Validate trained index. - Validate filled index. Args
Megatron-LM/megatron/core/datasets/retro/index/validate.py:179
Functionvalidate_latencies
Validate that latency decreases as we increase the number of cuda graphs. *Note*: This test is disabled for now, since the latency difference bet
Megatron-LM/tests/functional_tests/test_cases/gpt/gpt_dynamic_inference_tp1_pp1_583m_cuda_graphs_validation/cuda_graphs.py:173
Functionvalidate_loaded_state_dict
Validate the loaded state dict against the expected structure and types.
Megatron-LM/megatron/core/transformer/fsdp_dtensor_checkpoint.py:403
Methodvalidate_metadata_integrity
No reasonable checks can be applied
Megatron-LM/megatron/core/dist_checkpointing/mapping.py:654
Methodvalidate_recursive
(current_node, parent_weight_version)
slime/slime/router/middleware_hub/radix_tree.py:510
Methodvalidate_token_logp_consistency
Validate that token_ids, logp, and loss_mask have consistent lengths.
slime/slime/router/middleware_hub/radix_tree.py:66
Functionvalidator
Validates the request to the experimental function. Args: func (Callable): Callee max_lifetime (int, optional): Numbe
Megatron-LM/megatron/core/utils.py:120
Functionvalue_loss_function
Compute clipped value loss and metrics. Extracts current value predictions from `logits`, compares them against stored old values with clippi
slime/slime/backends/megatron_utils/loss.py:1078
Functionvanilla_tis_function
( args, *, pg_loss: torch.Tensor, train_log_probs: list[torch.Tensor], rollout_log_probs:
slime/slime/backends/megatron_utils/loss.py:795
Methodvm_platform
(self)
gui-rl/desktop_env/desktop_env.py:317
Methodvm_screen_size
(self)
gui-rl/desktop_env/desktop_env.py:321
Methodvocab
Returns tokenizer vocab.
Megatron-LM/megatron/core/tokenizers/base_tokenizer.py:36
Methodvocab
Returns tokenizer vocabulary.
Megatron-LM/megatron/core/tokenizers/text/text_tokenizer.py:155
Methodvocab
Megatron-LM/megatron/core/tokenizers/text/libraries/null_tokenizer.py:55
Methodvocab
Returns tokenizer vocab values.
Megatron-LM/megatron/core/tokenizers/text/libraries/huggingface_tokenizer.py:261
Methodvocab
Returns tokenizer's vocabulary.
Megatron-LM/megatron/core/tokenizers/text/libraries/sentencepiece_tokenizer.py:397
Methodvocab
Returns tokenizer vocab.
Megatron-LM/megatron/core/tokenizers/text/libraries/tiktoken_tokenizer.py:281
Methodvocab
Dictionary from vocab text token to id token
Megatron-LM/megatron/core/datasets/megatron_tokenizer.py:85
Methodvocab
Dictionary from vocab text token to id token.
Megatron-LM/megatron/training/tokenizer/tokenizer.py:150
Methodvocab
(self)
Megatron-LM/megatron/training/tokenizer/tokenizer.py:241
Methodvocab
(self)
Megatron-LM/megatron/training/tokenizer/tokenizer.py:346
Methodvocab
(self)
Megatron-LM/megatron/training/tokenizer/tokenizer.py:446
Methodvocab
(self)
Megatron-LM/megatron/training/tokenizer/tokenizer.py:769
← previousnext →10,401–10,500 of 10,561, ranked by callers