Code
Hub
Workspaces
Following
Trending
Connect
MCP
copy
Create free account
hub
/
github.com/Gen-Verse/OpenClaw-RL
/ functions
Functions
10,561 in github.com/Gen-Verse/OpenClaw-RL
⨍
Functions
10,561
◇
Types & classes
1,405
↳
Endpoints
144
↓ 2 callers
Function
_defer_loading_sharded_items
Divides state dict into parts loaded by this vs other ranks. Args: sharded_state_dict (ShardedStateDict): state dict with sharded items
Megatron-LM/megatron/core/dist_checkpointing/strategies/fully_parallel.py:465
↓ 2 callers
Method
_delete_vm
(self, vm_path)
gui-rl/desktop_env/providers/virtualbox/manager.py:363
↓ 2 callers
Method
_delete_vm
(self, vm_path)
gui-rl/desktop_env/providers/vmware/manager.py:351
↓ 2 callers
Method
_download
Gets a file from cache or downloads it Args: path: path to the file in cache url: url to the file Re
Megatron-LM/megatron/core/tokenizers/text/libraries/megatron_hf_tokenizer.py:149
↓ 2 callers
Method
_dynamic_step_context_init
Initializes the inference context for dynamic batching. Args: construct_graph_dimensions (Optional[InferenceBatchDimensions]): Th
Megatron-LM/megatron/core/inference/text_generation_controllers/text_generation_controller.py:484
↓ 2 callers
Method
_dynamic_step_forward_logits
Forward step the model to get logits for dynamic batching. This also handles logits-broadcasting for pipeline parallelism. Args:
Megatron-LM/megatron/core/inference/text_generation_controllers/text_generation_controller.py:538
↓ 2 callers
Method
_dynamic_step_sample_logits
Sample tokens from logits for dynamic batching. Args: logits (Tensor): The logits to sample from. backend (str): The
Megatron-LM/megatron/core/inference/text_generation_controllers/text_generation_controller.py:656
↓ 2 callers
Method
_encode_extra_state
(self, state)
Megatron-LM/megatron/core/extensions/transformer_engine.py:1435
↓ 2 callers
Method
_enforce_double_buffer_limit
(self, add_buckets)
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:3028
↓ 2 callers
Function
_ensure_qwen35_attention_layer_ids
(config)
slime/slime/backends/sglang_utils/qwen3_5.py:377
↓ 2 callers
Function
_eps_clip_hi
(args: Namespace)
openclaw-combine/hint_opd_loss.py:239
↓ 2 callers
Function
_eps_clip_lo
(args: Namespace)
openclaw-combine/hint_opd_loss.py:234
↓ 2 callers
Method
_evaluation
( self, prompt: Any, golden: dict | None, request: EvaluationRequest )
Megatron-LM/megatron/rl/agent/pass_at_evaluation_agent.py:54
↓ 2 callers
Method
_expected_shape
(sh_ten)
Megatron-LM/megatron/core/dist_checkpointing/strategies/torch.py:590
↓ 2 callers
Function
_extract_bullets
(xml_data)
gui-rl/desktop_env/evaluators/metrics/slides.py:598
↓ 2 callers
Method
_extract_common_per_param_step
(state_dict)
Megatron-LM/megatron/core/optimizer/optimizer.py:325
↓ 2 callers
Method
_extract_from_cache
Extract `size` bytes starting at `offset` bytes into the cache
Megatron-LM/megatron/core/datasets/indexed_dataset.py:475
↓ 2 callers
Method
_extract_multimodal
(messages: List[Dict[str, Any]], processor: Any)
gui-rl/agents/qwen35_agent.py:278
↓ 2 callers
Method
_extract_multimodal
(messages: List[Dict[str, Any]], processor: Any)
gui-rl/agents/qwen3vl_agent.py:228
↓ 2 callers
Function
_fill_in_deferred_sharded_items
Helper function to fill in items not loaded by current rank.
Megatron-LM/megatron/core/dist_checkpointing/strategies/fully_parallel.py:502
↓ 2 callers
Method
_filter_and_reorder_param_groups
Filter and reorder state_dict parameter groups to match current optimizer groups. Keys used for matching align with those from _get_param_grou
Megatron-LM/megatron/core/optimizer/optimizer.py:383
↓ 2 callers
Function
_filter_optimizer_tensors
Return only optimizer-related tensors from a flat checkpoint tensor dict.
Megatron-LM/tests/functional_tests/python_test_utils/test_optimizer_grads_match.py:104
↓ 2 callers
Function
_fit
(lst: list[float], length: int)
openclaw-tinker/data_formatter.py:54
↓ 2 callers
Function
_fix_dtype
(d)
slime/slime_plugins/models/hf_attention.py:34
↓ 2 callers
Function
_fix_pyautogui_less_than_bug
Fix PyAutoGUI '<' character bug by converting it to hotkey("shift", ',') calls. This fixes the known PyAutoGUI issue where typing '<' pr
gui-rl/desktop_env/desktop_env.py:26
↓ 2 callers
Method
_flush_pending_record
Write out the buffered record for *session_id* with its next_state and fire PRM.
openclaw-rl/openclaw_api_server.py:315
↓ 2 callers
Method
_flush_pending_record
(self, session_id, next_state)
openclaw-rl/oel/openclaw_oel_api_server.py:1526
↓ 2 callers
Method
_flush_pending_record
(self, session_id: str, next_state: dict[str, Any] | None)
openclaw-opd/openclaw_opd_api_server.py:395
↓ 2 callers
Method
_flush_pending_record
(self, session_id, next_state)
openclaw-fireworks/fireworks_server.py:844
↓ 2 callers
Method
_flush_pending_record
(self, session_id: str, next_state)
openclaw-tinker/api_server.py:494
↓ 2 callers
Function
_fork_rng
Fork the rng state.
Megatron-LM/megatron/core/tensor_parallel/random.py:464
↓ 2 callers
Method
_format_and_extract_experience
Format a conversation and extract experience items via PRM. Returns parsed experience text, or ``"No previous experience."`` if extra
openclaw-rl/oel/openclaw_oel_api_server.py:907
↓ 2 callers
Method
_format_output
De-duplicate and combine output_head + output_tail.
swe-rl/swe_prm.py:155
↓ 2 callers
Method
_forward_post_mlp
Perform operations after the MLP computation. Args: mlp_output_with_bias (Tensor): Output tensor of the MLP layer with b
Megatron-LM/megatron/core/transformer/transformer_layer.py:714
↓ 2 callers
Function
_free_storage
Frees the underlying storage of ``tensor``. Returns: bool: ``True`` if the method freed the storage and ``False`` if the sto
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:135
↓ 2 callers
Method
_fsdp2_load_full_state_dict
Load full state dict into FSDP2 model with efficient broadcast from rank 0. This function loads weights from rank 0 and broadcasts to all oth
slime/slime/backends/fsdp_utils/actor.py:259
↓ 2 callers
Function
_gather_from_tensor_parallel_region
(x: Tensor, config: TransformerConfig)
Megatron-LM/megatron/core/transformer/heterogeneous/linear_replacements.py:28
↓ 2 callers
Function
_gather_teacher_at_student_topk
Gather teacher log-probs at student's top-K token positions. The teacher's top-N (N >= K, typically N=1000 when topk_source=student) is used
openclaw-rl/oel/oel_distillation_loss.py:74
↓ 2 callers
Function
_generate_data_iterator
(rollout_data, micro_batch_size, micro_batch_indices=None)
slime/slime/backends/megatron_utils/data.py:360
↓ 2 callers
Method
_generate_experience_text
Use PRM engine to generate experience text.
openclaw-rl/oel/openclaw_oel_api_server.py:1175
↓ 2 callers
Method
_get
(cls, schema, model)
Megatron-LM/tools/checkpoint/schema_base.py:62
↓ 2 callers
Function
_get_artifact_name_and_version
(save_dir: Path, checkpoint_path: Path)
Megatron-LM/megatron/training/wandb_utils.py:15
↓ 2 callers
Method
_get_cuda_graph_input_data
Create the CUDA Graph capturing input data. The data is organized per-chunk per-microbatch per-layer.
Megatron-LM/megatron/core/transformer/cuda_graphs.py:1740
↓ 2 callers
Function
_get_custom_recipe
(quantizer_factory_python_path: str)
Megatron-LM/megatron/core/fp8_utils.py:155
↓ 2 callers
Method
_get_distribution
( fully_parallel, sharded_part, parallelization_group, cached_distribution=None )
Megatron-LM/megatron/core/dist_checkpointing/tensor_aware_state_dict.py:68
↓ 2 callers
Function
_get_dp_tp_mesh
(dp_cp_group, tp_group, ep_size=1)
Megatron-LM/megatron/core/distributed/fsdp/mcore_fsdp_adapter.py:363
↓ 2 callers
Method
_get_elapsed_time_all_ranks
Returns elapsed times of timers in names. Assumptions: - All the ranks call this function. - `names` are identical on
Megatron-LM/megatron/core/timers.py:259
↓ 2 callers
Method
_get_embeddings
Preprocesses input data for the Multi-Token Prediction (MTP) layers. This function computes the decoder input and sends updated inpu
Megatron-LM/megatron/core/transformer/multi_token_prediction.py:687
↓ 2 callers
Function
_get_empty_tensor_for_exchange
Determines the empty tensor to use for exchange. If shard_id is needed by this rank, it will be in the `unloaded_shards`. Otherwise, the meta
Megatron-LM/megatron/core/dist_checkpointing/exchange_utils.py:72
↓ 2 callers
Method
_get_eta_text
Calculate estimated time remaining based on current progress.
swe-rl/mini-swe-agent/src/minisweagent/run/extra/utils/batch_progress.py:89
↓ 2 callers
Function
_get_field
(string, type)
Megatron-LM/megatron/training/training.py:557
↓ 2 callers
Method
_get_gbuf_name
(self, buf_group_id: int, bucket_index: int)
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:769
↓ 2 callers
Method
_get_global_min_max_time
Report only min and max times across all ranks.
Megatron-LM/megatron/core/timers.py:307
↓ 2 callers
Function
_get_gui_trajectory_semaphore
()
gui-rl/generate_with_gui.py:44
↓ 2 callers
Method
_get_init_weight_context_manager
Get context manager for model initialization. Returns a callable that creates a context manager. Uses meta device (no memory allocati
slime/slime/backends/fsdp_utils/actor.py:234
↓ 2 callers
Method
_get_item_local_index
Return the local coordinates of the slice of this buffer's data that contains the item with the given ID.
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:1086
↓ 2 callers
Method
_get_item_local_shard_index
Return the local coordinates of the slice of this buffer's shard that contains the item with the given ID. In other words, this retur
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:1042
↓ 2 callers
Function
_get_keys_endswith
Retrieve keys from the model that end with a specified suffix.
Megatron-LM/megatron/core/transformer/moe/upcycling_utils.py:15
↓ 2 callers
Function
_get_layer_norm
(config: AttentionConfig | MLPConfig, use_te: bool, normalization: str)
Megatron-LM/megatron/core/models/gpt/heterogeneous/heterogeneous_layer_specs.py:68
↓ 2 callers
Function
_get_lock_path
(path_pattern: str, gpu_id: int)
slime/tests/ci/gpu_lock_exec.py:186
↓ 2 callers
Function
_get_ltor_masks_and_position_ids
Build masks and position id for left to right model. Args: data (torch.Tensor): The data tenor that holds the tokens from the dataset
Megatron-LM/megatron/core/datasets/gpt_dataset.py:630
↓ 2 callers
Function
_get_metadata_path
Returns metadata file path. Args: tokenizer_path (str): path to the tokenizer model. Returns: str: path to the metadata
Megatron-LM/megatron/core/tokenizers/megatron_tokenizer.py:156
↓ 2 callers
Function
_get_mlp_module_spec
(use_te: bool = True)
Megatron-LM/megatron/core/models/vision/vit_layer_specs.py:88
↓ 2 callers
Method
_get_model_and_main_params_data_float16
(self)
Megatron-LM/megatron/core/optimizer/optimizer.py:745
↓ 2 callers
Method
_get_model_inputs_args
(self, packed_sequence: dict)
slime/slime/backends/fsdp_utils/actor.py:841
↓ 2 callers
Method
_get_overview_data
Get data like exit statuses, total costs, etc.
swe-rl/mini-swe-agent/src/minisweagent/run/extra/utils/batch_progress.py:168
↓ 2 callers
Function
_get_param_groups
Create parameter groups for optimizer. Creates parameter groups from provided optimizer config object. Args: model_chunks (List[Mega
Megatron-LM/megatron/core/optimizer/__init__.py:91
↓ 2 callers
Function
_get_platform
Helper function to lazy initialize the platform.
Megatron-LM/megatron/core/fusions/fused_linear_cross_entropy.py:46
↓ 2 callers
Function
_get_qwen35_layer_spec
(provider: "Qwen35VLModelProvider", vp_stage=None)
slime/slime_plugins/megatron_bridge/qwen35_vl.py:845
↓ 2 callers
Function
_get_test_logger
Return the same logger instance used inside mcore utils.
Megatron-LM/tests/unit_tests/utils/test_experimental_log_once.py:12
↓ 2 callers
Function
_get_text_config
Extract text config from a VLM config if needed.
slime/slime_plugins/models/qwen3_5.py:21
↓ 2 callers
Function
_get_thd_freqs_on_this_cp_rank
Get the correct frequency slice for this context parallel rank with optional sequence offset. Args: cp_rank: Current context parallel ran
Megatron-LM/megatron/core/models/common/embeddings/rope_utils.py:129
↓ 2 callers
Method
_get_token_delta
Calculate token delta for multi-turn conversations. Tokenization logic adapted from: https://verl.readthedocs.io/en/v0.4.1/s
slime/examples/tau-bench/trainable_agents.py:334
↓ 2 callers
Method
_get_trtllm_config
Get TRTLLM Config Returns appropriate TRTLLM PretrainedConfig used by TRTLLM for building engine Args: export_config (Ex
Megatron-LM/megatron/core/export/trtllm/trtllm_helper.py:110
↓ 2 callers
Method
_get_valid_groups_with_timeout
Get valid groups including timeout-based groups
slime/slime_plugins/rollout_buffer/buffer.py:162
↓ 2 callers
Function
_get_wandb_artifact_tracker_filename
Wandb artifact tracker file records the latest artifact wandb entity and project
Megatron-LM/megatron/training/wandb_utils.py:10
↓ 2 callers
Function
_get_weight_scale
()
Megatron-LM/megatron/core/transformer/moe/upcycling_utils.py:155
↓ 2 callers
Function
_infer_instance_type
(instance: dict)
swe-rl/generate_with_swe_trajectory.py:319
↓ 2 callers
Method
_init_dynamic_sampling_tensors
Initialize tensors needed for dynamic sampling.
Megatron-LM/megatron/core/inference/text_generation_controllers/text_generation_controller.py:85
↓ 2 callers
Method
_init_sub_optimizers
(self)
Megatron-LM/megatron/core/optimizer/cpu_offloading/hybrid_optimizer.py:181
↓ 2 callers
Function
_init_wandb_common
()
slime/slime/utils/wandb_utils.py:150
↓ 2 callers
Function
_initialize_tp_communicators
initializing the communicators with user buffers for high-performance tensor-model-parallel communication overlap
Megatron-LM/megatron/training/initialize.py:238
↓ 2 callers
Function
_is_cuda_contiguous
Check if a tensor is not none, is cuda, and is contiguous.
Megatron-LM/megatron/core/inference/communication_utils.py:34
↓ 2 callers
Function
_is_flow_limited
(e: Exception)
gui-rl/desktop_env/providers/volcengine/manager.py:102
↓ 2 callers
Function
_is_frac
(expr: str)
slime/slime/rollout/rm_hub/math_utils.py:210
↓ 2 callers
Function
_is_loopback
(ip)
slime/slime/utils/http_utils.py:48
↓ 2 callers
Function
_is_offline_mode
Detect whether W&B should run in offline mode. Priority order: 1) args.wandb_mode if provided 2) WANDB_MODE environment variable
slime/slime/utils/wandb_utils.py:10
↓ 2 callers
Method
_is_proxy_available
Check if proxy is available
gui-rl/desktop_env/providers/aws/proxy_pool.py:84
↓ 2 callers
Method
_is_raisable
(value: Any)
Megatron-LM/megatron/core/inference/async_stream.py:70
↓ 2 callers
Function
_is_valid_git_patch
(patch_text: str)
swe-rl/generate_with_swe_trajectory.py:197
↓ 2 callers
Function
_kernel_make_viewless_tensor
Make a viewless tensor. View tensors have the undesirable side-affect of retaining a reference to the originally-viewed tensor, even after ma
Megatron-LM/megatron/core/utils.py:632
↓ 2 callers
Function
_kill_container_processes
Kill all user processes inside a container after a timeout. With --init, tini is PID 1 and will reap children. This sends SIGKILL to all rem
swe-rl/server/swe_exec_server.py:60
↓ 2 callers
Function
_layer_is_graphable
Check if a layer is graphable.
Megatron-LM/megatron/core/transformer/cuda_graphs.py:1350
↓ 2 callers
Function
_load_checkpoint
(queue, args)
Megatron-LM/tools/checkpoint/loader_legacy.py:34
↓ 2 callers
Function
_load_global_dist_base_checkpoint
Load the base state_dict from the given directory containing the global distributed checkpoint
Megatron-LM/megatron/training/checkpointing.py:1038
↓ 2 callers
Function
_load_hf_config
Load HF config with a JSON fallback for unsupported model types. Some Qwen3.5 / Qwen3.5-VL checkpoints carry ``model_type`` strings that upst
slime/slime_plugins/models/hf_attention.py:14
↓ 2 callers
Function
_load_meta_pairs
(meta_path: str)
gui-rl/generate_with_gui.py:61
↓ 2 callers
Function
_load_task_config
(base_dir: str, domain: str, example_id: str)
gui-rl/generate_with_gui.py:79
↓ 2 callers
Function
_log_verbose
(name: str, start: float)
Megatron-LM/megatron/core/utils.py:2303
↓ 2 callers
Method
_make_backward_post_hook
Creates a backward post-hook to dispatch an all-reduce / reduce-scatter when ready (i.e., when all grads in a bucket have been comput
Megatron-LM/megatron/core/distributed/distributed_data_parallel.py:419
← previous
next →
1,501–1,600 of 10,561, ranked by callers