MCPcopy Create free account

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

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

↓ 1 callersFunction_collate_fn
Collate function for the DataLoader. Args: batch: List of dictionaries from the dataset Returns: Dictionary of batched
Megatron-LM/examples/mimo/data/mock.py:211
↓ 1 callersMethod_collect_main_grad_data_for_unscaling
(self)
Megatron-LM/megatron/core/optimizer/optimizer.py:725
↓ 1 callersMethod_collect_original_tensor_info
(self)
Megatron-LM/megatron/core/post_training/modelopt/layers.py:227
↓ 1 callersMethod_communicate_shapes
Communicate tensor shapes between stages. Used to communicate tensor shapes before the actual tensor communication happens. This is re
Megatron-LM/megatron/core/pipeline_parallel/p2p_communication.py:165
↓ 1 callersFunction_compare_bullets_with_tolerance
Compare bullets with tolerance for minor differences
gui-rl/desktop_env/evaluators/metrics/slides.py:629
↓ 1 callersFunction_compare_dataclasses
(obj1, obj2)
Megatron-LM/megatron/core/dist_checkpointing/strategies/state_dict_saver.py:27
↓ 1 callersFunction_compare_two_checkpoint
(checkpoint_1, checkpoint_2)
Megatron-LM/tools/checkpoint/checkpoint_inspector.py:870
↓ 1 callersFunction_compile_dependencies
()
Megatron-LM/megatron/training/initialize.py:172
↓ 1 callersFunction_compile_timeout
Context manager to timeout compilation. Args: timeout_s (int): Timeout in seconds.
Megatron-LM/megatron/core/inference/unified_memory.py:48
↓ 1 callersFunction_compose_plugin_maybe_missing
(stderr: str | None)
terminal-rl/remote/docker_compose_utils.py:95
↓ 1 callersFunction_compute_attention_layer_ids
(config)
slime/slime/backends/sglang_utils/qwen3_5.py:365
↓ 1 callersMethod_compute_bias
Adapted from HuggingFace T5 Model https://github.com/huggingface/transformers/blob/329f5dbf97a5cb2473914c88c05aa3dcb242e19a/
Megatron-LM/megatron/core/models/common/embeddings/relative_pos_embedding.py:100
↓ 1 callersFunction_compute_config_for_logging
(args)
slime/slime/utils/wandb_utils.py:82
↓ 1 callersMethod_compute_dynamic_global_batch_size
Calculate dynamic global_batch_size from actual per-rollout samples. If target_steps is provided, choose global_batch_size to keep the realiz
slime/slime/ray/rollout.py:287
↓ 1 callersMethod_compute_grid
( problem_mnk: Tuple[int, int, int], cluster_shape_mn: Tuple[int, int], ct
Megatron-LM/megatron/core/fusions/linear_cross_entropy/blackwell/fwd_mainloop.py:532
↓ 1 callersMethod_compute_grid
( self, problem_mnk: Tuple[int, int, int], cluster_shape_mn: Tuple[int, in
Megatron-LM/megatron/core/fusions/linear_cross_entropy/blackwell/bwd_partial_dlogits.py:67
↓ 1 callersFunction_compute_hash_tensor
(x: torch.Tensor)
slime/slime/utils/tensor_backper.py:109
↓ 1 callersMethod_compute_index_scores
Perform index score using BF16 precision. Reference: https://github.com/deepseek-ai/DeepSeek-V3.2-Exp/blob/main/inferenc
Megatron-LM/megatron/core/transformer/experimental_attention_variant/dsa.py:474
↓ 1 callersFunction_compute_logprob_entropy_chunked
( logits: torch.Tensor, tokens: torch.Tensor, tp_group, with_entropy: bool, chunk_size: in
slime/slime/utils/ppo_utils.py:655
↓ 1 callersMethod_compute_prm_teacher_log_probs_multi_cand
K-loop teacher forward for the retool-hybrid-select path. ``teacher_tokens_cand[i]`` is a non-empty list of ``K_i`` LongTensors (each
slime/slime/backends/megatron_utils/actor.py:679
↓ 1 callersFunction_compute_reward_cat_metrics
(args, all_samples: list[Sample])
slime/slime/ray/rollout.py:1316
↓ 1 callersFunction_compute_server_args
( args, rank, dist_init_addr, nccl_port, host, port, worker_type: str = "regular",
slime/slime/backends/sglang_utils/sglang_engine.py:491
↓ 1 callersFunction_compute_shards_access
(rank_sharding)
Megatron-LM/megatron/core/dist_checkpointing/validation.py:500
↓ 1 callersMethod_compute_stages
( self, tiled_mma: cute.TiledMma, mma_tiler: Tuple[int, int, int],
Megatron-LM/megatron/core/fusions/linear_cross_entropy/blackwell/fwd_mainloop.py:88
↓ 1 callersMethod_compute_stages
( self, tiled_mma: cute.TiledMma, mma_tiler: Tuple[int, int, int],
Megatron-LM/megatron/core/fusions/linear_cross_entropy/blackwell/bwd_partial_dlogits.py:85
↓ 1 callersMethod_compute_teacher_log_probs
(self, input_ids: list[int], response_len: int)
openclaw-rl/oel/openclaw_oel_api_server.py:1251
↓ 1 callersMethod_compute_teacher_log_probs
(self, input_ids: list[int], response_len: int)
openclaw-opd/openclaw_opd_api_server.py:476
↓ 1 callersMethod_compute_teacher_log_probs
Compute teacher log-probs using the Fireworks policy trainer. Builds a datum from the hint-augmented full sequence and runs a forward
openclaw-fireworks/fireworks_server.py:467
↓ 1 callersMethod_compute_teacher_topk_logprobs
( self, input_ids: list[int], response_len: int )
openclaw-rl/oel/openclaw_oel_api_server.py:1290
↓ 1 callersMethod_compute_teacher_topk_logprobs
Compute teacher's top-K log-probs and token indices for response tokens. Returns: (logprobs, indices) where each is a list of len
openclaw-opd/openclaw_opd_api_server.py:515
↓ 1 callersFunction_compute_wrapper_tokens
Compute the (small, deterministic) wrapper-token sequences the chat template inserts. Cached per-tokenizer-id to avoid repeated encoding. Ret
swe-rl/generate_with_swe_trajectory.py:470
↓ 1 callersFunction_compute_zero_std_metrics
(args, all_samples: list[Sample])
slime/slime/ray/rollout.py:1278
↓ 1 callersMethod_connect_with_inference_coordinator
Performs the initial handshake with the inference coordinator. Sends a CONNECT signal and waits for a CONNECT_ACK reply to ensure th
Megatron-LM/megatron/core/inference/inference_client.py:155
↓ 1 callersMethod_controller
Installs a controller listener that is used to toggle collection state. Called from configure(). Ignored for all ranks other than rank-0
Megatron-LM/megatron/core/utils.py:1705
↓ 1 callersFunction_convert_bbox_format_from_abs_origin
(bbox, image_ele: dict, *, tgt_format: str)
gui-rl/agents/utils/qwen_vl_utils.py:84
↓ 1 callersFunction_convert_bbox_format_to_abs_origin
(bbox, image_ele: dict, *, src_format: str)
gui-rl/agents/utils/qwen_vl_utils.py:121
↓ 1 callersFunction_convert_fsdp_to_hf
( origin_hf_dir: str, input_dir: str, output_dir: str, )
slime/tools/convert_fsdp_to_hf.py:115
↓ 1 callersFunction_convert_mtp_layer
Convert MTP layer parameters from Megatron to HuggingFace format. Handles both wrapper layers (enorm, hnorm, final_layernorm, eh_proj) and in
slime/slime/backends/megatron_utils/megatron_to_hf/qwen3_next.py:6
↓ 1 callersFunction_convert_mtp_layer
(args, name, param, layer_idx)
slime/slime/backends/megatron_utils/megatron_to_hf/qwen3_5.py:77
↓ 1 callersMethod_convert_mtp_param
Convert MTP layer parameters from MCore to HF format.
slime/slime_plugins/mbridge/qwen3_next.py:58
↓ 1 callersMethod_convert_mtp_param
Convert MTP layer parameters from MCore to HF format.
slime/slime_plugins/mbridge/mimo.py:58
↓ 1 callersMethod_convert_non_transformer_layer
Convert Non Transformer layers to TRTLLM weights Non transformer layers referes to layers that occur only once in the model (e.g Embedding ,
Megatron-LM/megatron/core/export/trtllm/trtllm_weights_converter/single_device_trtllm_model_weights_converter.py:81
↓ 1 callersMethod_convert_non_transformer_layer
Convert Non Transformer layers to TRTLLM weights Non transformer layers referes to layers that occur only once in the model (e.g Embedding ,
Megatron-LM/megatron/core/export/trtllm/trtllm_weights_converter/distributed_trtllm_model_weights_converter.py:195
↓ 1 callersFunction_convert_point_format_from_abs_origin
(point, image_ele: dict, *, tgt_format: str)
gui-rl/agents/utils/qwen_vl_utils.py:164
↓ 1 callersFunction_convert_point_format_to_abs_origin
(point, image_ele: dict, *, src_format: str)
gui-rl/agents/utils/qwen_vl_utils.py:193
↓ 1 callersMethod_convert_samples_to_train_data
Convert inference generated samples to training data.
slime/slime/ray/rollout.py:602
↓ 1 callersFunction_convert_to_hf_core
(args, model_name, name, param)
slime/slime/backends/megatron_utils/megatron_to_hf/__init__.py:35
↓ 1 callersMethod_convert_to_hf_named_tensors
(self, megatron_full_params: Sequence[torch.Tensor], param_infos: list[ParamInfo])
slime/slime/backends/megatron_utils/update_weight/hf_weight_iterator_direct.py:35
↓ 1 callersMethod_convert_to_openai_message
Convert parsing results to OpenAI format assistant message Args: normal_text: Normal text content calls: Lis
slime/examples/tau-bench/openai_tool_adapter.py:81
↓ 1 callersMethod_convert_transformer_layer
Convert Transformer layers to TRTLLM weights Transformer layers referes to layers within the transformber block. They have a layer number ass
Megatron-LM/megatron/core/export/trtllm/trtllm_weights_converter/single_device_trtllm_model_weights_converter.py:114
↓ 1 callersMethod_convert_transformer_layer
Convert Transformer layers to TRTLLM weights Transformer layers referes to layers within the transformber block. They have a layer number ass
Megatron-LM/megatron/core/export/trtllm/trtllm_weights_converter/distributed_trtllm_model_weights_converter.py:100
↓ 1 callersFunction_convert_vision_model_to_hf
(name, param)
slime/slime/backends/megatron_utils/megatron_to_hf/qwen3_5.py:35
↓ 1 callersMethod_copy_main_params_to_model_params
(self)
Megatron-LM/megatron/core/optimizer/optimizer.py:774
↓ 1 callersMethod_copy_model_grads_to_main_grads
(self)
Megatron-LM/megatron/core/optimizer/optimizer.py:754
↓ 1 callersMethod_copy_model_params_to_main_params
(self, state_dict=None)
Megatron-LM/megatron/core/optimizer/optimizer.py:781
↓ 1 callersFunction_correct_amax_history_if_needed_impl
(model: List[torch.nn.Module])
Megatron-LM/megatron/core/fp8_utils.py:267
↓ 1 callersFunction_create_agentnet_agent
(args: Any, *, max_steps: int, max_image_history_length: int, result_dir: Path)
gui-rl/generate_with_agentnet.py:41
↓ 1 callersMethod_create_and_store_slot
(self, idx: int, count: int)
gui-rl/env_pool_server.py:78
↓ 1 callersFunction_create_atspi_node
(node: Accessible, depth: int = 0, flag: Optional[str] = None)
gui-rl/desktop_env/server/main.py:423
↓ 1 callersMethod_create_batch
Create a simple batch with mixed text and audio content.
Megatron-LM/tests/unit_tests/models/test_mimo_audio_submodules.py:283
↓ 1 callersFunction_create_build_dir
(buildpath)
Megatron-LM/megatron/legacy/fused_kernels/__init__.py:70
↓ 1 callersMethod_create_client
(self)
gui-rl/desktop_env/providers/volcengine/provider.py:26
↓ 1 callersMethod_create_client
(self)
gui-rl/desktop_env/providers/aliyun/provider.py:33
↓ 1 callersFunction_create_gui_agent
(args: Any, *, max_steps: int, max_image_history_length: int, result_dir: Path)
gui-rl/generate_with_gui.py:199
↓ 1 callersFunction_create_or_open_zarr_arrays
Returns list of zarr arrays corresponding to given tensors. For a sharded tensors that: a) is main replica and represents the first chunk (al
Megatron-LM/megatron/core/dist_checkpointing/strategies/zarr.py:96
↓ 1 callersFunction_create_placement_group
Create a placement group with the specified number of GPUs.
slime/slime/ray/placement_group.py:41
↓ 1 callersMethod_create_ref_model
Create and initialize a separate reference model with FSDP2 CPUOffloadPolicy. Parameters: ref_load_path: Path to a directory cont
slime/slime/backends/fsdp_utils/actor.py:800
↓ 1 callersMethod_create_safe_environment
Create safe execution environment with temporary directory
toolcall-rl/tool_sandbox.py:168
↓ 1 callersMethod_create_sample_audio
Create a sample audio waveform. Args: duration_seconds (float): Duration of audio in seconds sample_rate (int, option
Megatron-LM/tests/unit_tests/models/test_mimo_audio_submodules.py:200
↓ 1 callersFunction_create_zarr_array
(sharded_tensor: ShardedTensor, checkpoint_dir: Path)
Megatron-LM/megatron/core/dist_checkpointing/strategies/zarr.py:167
↓ 1 callersFunction_custom_prompt_forward_loop_func
(model)
Megatron-LM/examples/post_training/modelopt/prune.py:191
↓ 1 callersFunction_custom_prompt_forward_loop_func
(model)
Megatron-LM/examples/post_training/modelopt/quantize.py:181
↓ 1 callersMethod_decode_extra_state
(self, state)
Megatron-LM/megatron/core/extensions/kitchen.py:689
↓ 1 callersMethod_decode_text
Decode exactly like policy: output_token_logprobs -> item[1] FIRST, fall back to output["text"] only if logprobs are absent.
gui-rl/agents/qwen3vl_reward_agent.py:254
↓ 1 callersMethod_decode_text
(self, output: Any)
swe-rl/swe_prm.py:261
↓ 1 callersFunction_decorate
(fn: Callable)
Megatron-LM/megatron/core/utils.py:2317
↓ 1 callersMethod_defer_loading_sharded_objects
( sharded_state_dict: ShardedStateDict, )
Megatron-LM/megatron/core/dist_checkpointing/strategies/fully_parallel.py:293
↓ 1 callersMethod_defer_loading_sharded_tensors
( sharded_state_dict: ShardedStateDict, )
Megatron-LM/megatron/core/dist_checkpointing/strategies/fully_parallel.py:304
↓ 1 callersFunction_dequantize_tensor
Unpack int32 quantized tensor and multiply with scales to create BF16 tensor.
slime/tools/convert_k2_thinking_int4_to_bf16.py:55
↓ 1 callersFunction_derive_status
(outcome: RolloutOutcome, eval_error: str | None)
terminal-rl/sample_builders.py:14
↓ 1 callersFunction_detect_model_dir
(input_dir: str)
slime/tools/convert_fsdp_to_hf.py:57
↓ 1 callersFunction_detect_tool_call_format
Detect whether the model uses JSON or XML tool call format. Qwen3.5 chat template uses '<function=' XML format; Qwen3 and others use JSON '{"
toolcall-rl/generate_with_retool.py:130
↓ 1 callersFunction_determine_if_first_last_layer_of_this_vp_chunk
Determine if the given module is the first/last layer of the PP+VPP chunk it belongs to. Returns a tuple of two booleans indicating if the module
Megatron-LM/megatron/core/transformer/cuda_graphs.py:153
↓ 1 callersFunction_determine_if_transformer_decoder_layer
Determine if the given module is a transformer decoder layer.
Megatron-LM/megatron/core/transformer/cuda_graphs.py:132
↓ 1 callersFunction_determine_missing_and_unexpected_keys
Determines load mismatches based on metadata. There is an asymmetry between "unexpected" and "missing" keys. Unexpected keys can be determine
Megatron-LM/megatron/core/dist_checkpointing/validation.py:283
↓ 1 callersMethod_distribute_data_to_state_dict
( self, ten_meta: _ShardedTensorMetadata, loaded_ten: torch.Tensor, sharded_st
Megatron-LM/megatron/core/dist_checkpointing/strategies/two_stage.py:240
↓ 1 callersFunction_docker_image_exists_locally
(image: str, timeout: float = 30.0)
terminal-rl/remote/docker_compose_utils.py:148
↓ 1 callersFunction_download_vm
(vms_dir: str)
gui-rl/desktop_env/providers/docker/manager.py:31
↓ 1 callersFunction_drain_output_queue
(args, worker: AsyncRolloutWorker)
openclaw-rl/openclaw_rollout.py:89
↓ 1 callersFunction_drain_output_queue
(args, worker: AsyncRolloutWorker)
openclaw-rl/oel/openclaw_oel_rollout.py:102
↓ 1 callersFunction_drain_output_queue
(args, worker: AsyncRolloutWorker)
openclaw-opd/openclaw_opd_rollout.py:93
↓ 1 callersFunction_drain_output_queue
(args, worker: AsyncRolloutWorker)
openclaw-combine/openclaw_combine_rollout.py:93
↓ 1 callersFunction_drain_output_queue
(args, worker: AsyncRolloutWorker)
openclaw-combine/openclaw_combine_select_rollout.py:109
↓ 1 callersMethod_drain_prm_pending
( self, prm_pending: list[tuple[int, asyncio.Task]], prm_turn_scores: dict[int, float]
terminal-rl/agent_runner.py:271
↓ 1 callersMethod_drop_constant_reward_groups
Drop GRPO/GSPO groups whose rewards are all identical. Keep at least one group to avoid empty training data.
slime/slime/ray/rollout.py:400
↓ 1 callersFunction_dtype_size
Get the size of the dtype. Args: dtype (torch.dtype): The dtype to get the size of. Returns: int: The size of the dtype.
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:3751
↓ 1 callersMethod_dynamic_step_calculate_log_probs
Calculate log probs from logits.
Megatron-LM/megatron/core/inference/text_generation_controllers/text_generation_controller.py:739
↓ 1 callersMethod_dynamic_step_context_bookkeeping
Update the dynamic inference context after sampling. Return: Dict [str, Tensor]: A dictionary containing: active_
Megatron-LM/megatron/core/inference/text_generation_controllers/text_generation_controller.py:835
↓ 1 callersMethod_dynamic_step_log_probs_bookkeeping
Perform bookkeeping necessary to compute log probs for dynamic batching.
Megatron-LM/megatron/core/inference/text_generation_controllers/text_generation_controller.py:703
↓ 1 callersMethod_dynamic_step_top_n_logprobs_bookkeeping
Perform bookkeeping necessary to compute top-n log probs for dynamic batching.
Megatron-LM/megatron/core/inference/text_generation_controllers/text_generation_controller.py:720
← previousnext →2,701–2,800 of 10,561, ranked by callers