MCPcopy Create free account

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

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

Functionmodel_provider_llava_vlm
Build a LLaVA-style Vision-Language MIMO model composed of: • Vicuna language model. • CLIP ViT-L/14 vision encoder. • 2-layer MLP vi
Megatron-LM/examples/mimo/model_providers/llava_vlm.py:29
Functionmodel_provider_mock_vlm_single_encoder
Build a MIMO model with a vision encoder.
Megatron-LM/examples/mimo/model_providers/mock.py:28
Functionmodelopt_gpt_mamba_builder
Builds the model. Args: args (Namespace): The arguments namespace. pre_process (bool, optional): Set to true if you need to compu
Megatron-LM/megatron/post_training/model_builder.py:133
Methodmodify_logits_for_top_k_filtering
Set the logits for none top-k values to -inf.
Megatron-LM/megatron/core/inference/text_generation_controllers/text_generation_controller.py:249
Methodmodify_logits_for_top_p_filtering
Set the logits for none top-p values to -inf.
Megatron-LM/megatron/core/inference/text_generation_controllers/text_generation_controller.py:254
Functionmodify_state_dict
Modify state dict items in a Megatron Core Distributed Checkpoint.
Megatron-LM/tools/checkpoint/checkpoint_inspector.py:847
Functionmoe_freq_type
Frequency between MoE layers and Dense layers. Accepts either: - An integer N: Represents a 1:N ratio, meaning one expert layer for every N-1
Megatron-LM/megatron/training/arguments.py:312
Functionmtbench_to_oai_chat
Convert MTBench data to OpenAI chat completion format.
Megatron-LM/megatron/post_training/utils.py:30
Functionmtbench_to_oai_chat
Convert MTBench data to OpenAI chat completion format.
Megatron-LM/examples/post_training/modelopt/generate.py:51
Methodmul_reduce
(a, b)
slime/slime/utils/ppo_utils.py:168
Functionmulti_no_sync
()
Megatron-LM/megatron/core/pipeline_parallel/schedules.py:958
Functionmuon_init_state_fn
(opt, config=None)
Megatron-LM/megatron/core/optimizer/muon.py:265
Methodmy_rank
Can be called to get configured rank of this instance Returns: int: Configured rank for this instance
Megatron-LM/megatron/core/utils.py:1866
Methodn_calls
(self)
swe-rl/mini-swe-agent/src/minisweagent/models/__init__.py:38
Methodn_calls
(self)
swe-rl/mini-swe-agent/src/minisweagent/models/extra/roulette.py:27
Methodn_completed
(self)
swe-rl/mini-swe-agent/src/minisweagent/run/extra/utils/batch_progress.py:86
Methodn_steps
Number of steps in current trajectory.
swe-rl/mini-swe-agent/src/minisweagent/run/inspector.py:71
Methodn_trajectories
Number of trajectory files.
swe-rl/mini-swe-agent/src/minisweagent/run/inspector.py:90
Methodname
(self)
slime/slime/utils/reloadable_process_group.py:166
Methodneed_skip_prefetch
(bucket_id)
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:3419
Methodnested_sum
(x)
Megatron-LM/tests/unit_tests/dist_checkpointing/test_serialization.py:766
Functionnew_add_argument_wrapper
Add arguments to the parser, ensuring that the server arguments are prefixed and skippable.
slime/slime/backends/sglang_utils/arguments.py:60
Functionnew_function
(*args, **kwargs)
slime/slime/utils/reloadable_process_group.py:51
Functionnew_group
(*args, **kwargs)
slime/slime/utils/reloadable_process_group.py:27
Functionnew_init
(self, *args, **kwargs)
slime/slime/backends/megatron_utils/__init__.py:11
Methodnext_bucket_id
Search for the next bucket ID that is not in the list of all-gather buckets.
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:3396
Functionno_rope_freq_type
Controls which layers to skip performing Rotary Position Embedding. - An integer N: Represents a 1:N ratio, meaning RoPE is skipped every N-1 lay
Megatron-LM/megatron/training/arguments.py:289
Methodno_sync
Context manager that turns off gradient synchronization.
Megatron-LM/megatron/core/distributed/data_parallel_base.py:25
Methodno_sync
Context manager that turns off gradient synchronization.
Megatron-LM/megatron/core/distributed/distributed_data_parallel.py:448
Methodno_sync
Context manager that turns off gradient synchronization. For grads shard mode there will actually always be gradient sync happening.
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/megatron_fsdp.py:957
Methodno_weight_decay
(self)
Megatron-LM/megatron/legacy/model/vision/esvit_swin_backbone.py:590
Methodno_weight_decay
(self)
Megatron-LM/megatron/legacy/model/vision/swin_backbone.py:562
Methodno_weight_decay_keywords
(self)
Megatron-LM/megatron/legacy/model/vision/esvit_swin_backbone.py:594
Methodno_weight_decay_keywords
(self)
Megatron-LM/megatron/legacy/model/vision/swin_backbone.py:566
Functionnoop_decorator
No-op decorator
Megatron-LM/megatron/core/jit.py:11
Functionnormalize_group_data
(group, epsilon=1e-8, algo="grpo")
slime/slime_plugins/rollout_buffer/generator/base_generator.py:300
Methodnormalize_vals
(vals: torch.Tensor, std_normalization: bool = False)
slime/slime/ray/rollout.py:348
Functionnull_decorator
No-op decorator.
Megatron-LM/megatron/core/utils.py:84
Methodnull_method
Default method to initialize start/stop method ptrs
Megatron-LM/megatron/core/utils.py:1883
Methodnum_buckets
Return the number of buckets.
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:2595
Methodnum_buckets
Return the number of buckets.
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:2926
Methodnum_buckets
Return the number of buckets.
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:3315
Methodnum_decode_requests
Returns the number of decode requests.
Megatron-LM/megatron/core/inference/contexts/dynamic_context.py:1174
Methodnum_warmup_microbatches
Calculate the number of warmup microbatches for the current rank. Uses the same simple logic as P2PCommunicator: total_pipeline_stage
Megatron-LM/megatron/core/pipeline_parallel/multimodule_communicator.py:155
Methodnum_workers
(self)
terminal-rl/router_server.py:52
Methodnumel_low_level_dataset
(low_level_dataset: LowLevelDataset)
Megatron-LM/tools/retro/sft/dataset_conv.py:73
Methodnumel_low_level_dataset
Abstract method implementation Args: low_level_dataset (MockGPTLowLevelDataset): The underlying MockGPTLowLevelDataset R
Megatron-LM/megatron/core/datasets/gpt_dataset.py:801
Methodnumel_low_level_dataset
Return the number of elements in the underlying low level dataset for the purpose of segregating the train/valid/test split indices I
Megatron-LM/megatron/core/datasets/megatron_dataset.py:117
Methodnumel_low_level_dataset
(low_level_dataset: IndexedDataset)
Megatron-LM/megatron/core/datasets/masked_dataset.py:116
Methodnumel_low_level_dataset
(low_level_dataset: LowLevelDataset)
Megatron-LM/megatron/training/datasets/sft_dataset.py:61
Methodnumel_low_level_dataset
(low_level_dataset: LowLevelDataset)
Megatron-LM/tests/unit_tests/data/test_builder.py:80
Functionnvtx_decorator
Decorator to add NVTX range to a function. Args: message (str, optional): Custom message for the NVTX range. If None, uses function path
Megatron-LM/megatron/core/utils.py:2201
Methodoccupy_vm
(self, vm_path, pid, region=DEFAULT_REGION, lock_needed=True, **kwargs)
gui-rl/desktop_env/providers/aws/manager.py:214
Methodoccupy_vm
(self, vm_path, pid, region=None, **kwargs)
gui-rl/desktop_env/providers/docker/manager.py:112
Methodoccupy_vm
(self, vm_path, pid, lock_needed=True)
gui-rl/desktop_env/providers/virtualbox/manager.py:334
Methodoccupy_vm
(self, vm_path, pid, lock_needed=True, **kwargs)
gui-rl/desktop_env/providers/volcengine/manager.py:311
Methodoccupy_vm
(self, vm_path, pid, lock_needed=True)
gui-rl/desktop_env/providers/vmware/manager.py:322
Methodoccupy_vm
(self, vm_path, pid, lock_needed=True, **kwargs)
gui-rl/desktop_env/providers/aliyun/manager.py:294
Functionoel_distillation_loss_function
Compute OEL full KL(student || teacher) with top-K + tail trick. Mathematics: KL(student || teacher) = sum_x P_s(x) * [log P_s(x) - log P
openclaw-rl/oel/oel_distillation_loss.py:134
Methodoffload
(self)
slime/slime/ray/actor_group.py:130
Methodoffload_to_cpu
Function used for RL training. Move optimizer state tensors to CPU to free GPU memory during inference.
Megatron-LM/megatron/core/optimizer/optimizer.py:344
Methodoffsets
Returns offsets.
Megatron-LM/megatron/core/tokenizers/text/libraries/null_tokenizer.py:36
Methodoffsets
(self, ids: list[int], text: str)
Megatron-LM/megatron/training/tokenizer/tokenizer.py:169
Methodoffsets
(self, ids: list[int], text: str)
Megatron-LM/megatron/training/tokenizer/tokenizer.py:502
Methodoffsets
(self, ids: list[int], text: str)
Megatron-LM/megatron/training/tokenizer/tokenizer.py:788
Methodoffsets
(self, ids: list[int], text: str)
Megatron-LM/megatron/training/tokenizer/tokenizer.py:898
Methodold_round_impl
(after, multiple)
Megatron-LM/tests/unit_tests/test_training.py:49
Methodon_agent_finished
(self, exit_status: str, result: str)
swe-rl/mini-swe-agent/src/minisweagent/agents/interactive_textual.py:348
Functionon_checkpoint_loaded
Should be called after a checkpoint was loaded Args: is_local_chkpt (bool): true if it was a local checkpoint, false if global
Megatron-LM/megatron/training/ft_integration.py:191
Functionon_checkpointing_end
Should be called after each checkpoint-saving-related operation. Args: is_async_finalization (bool): true if called after an async checkp
Megatron-LM/megatron/training/ft_integration.py:174
Functionon_checkpointing_start
Should be called before each checkpoint-saving-related operation.
Megatron-LM/megatron/training/ft_integration.py:167
Functionon_eval_step_end
Should be called after each validation step
Megatron-LM/megatron/training/ft_integration.py:157
Functionon_eval_step_start
Should be called before each validation step
Megatron-LM/megatron/training/ft_integration.py:144
Methodon_get_saved_tensor
Hook called when autograd retrieves a saved tensor during backward pass. Returns the actual tensor (potentially reloading from CPU).
Megatron-LM/megatron/core/pipeline_parallel/fine_grained_activation_offload.py:200
Functionon_load_checkpoint_success
Function to be called after succesful loading of a checkpoint, for aggregation and logging it to W&B Args: checkpoint_path (str): path of
Megatron-LM/megatron/training/wandb_utils.py:44
Methodon_log_message_emitted
Handle log messages of warning level or higher by showing them as notifications.
swe-rl/mini-swe-agent/src/minisweagent/agents/interactive_textual.py:338
Methodon_mount
(self)
swe-rl/mini-swe-agent/src/minisweagent/agents/interactive_textual.py:312
Methodon_mount
(self)
swe-rl/mini-swe-agent/src/minisweagent/run/inspector.py:133
Functionon_pretrain_start
Function to be called at the start of pretrain function to track E2E meta data
Megatron-LM/megatron/training/one_logger_utils.py:265
Functionon_save_checkpoint_end
Function to be called after checkpointing ends Args: save_checkpoint_duration (float): duration of current save checkpoint process
Megatron-LM/megatron/training/one_logger_utils.py:373
Functionon_save_checkpoint_success
Function to be called after checkpointing succeeds and checkpoint is persisted for storing productive metrics Args: productive_metrics (d
Megatron-LM/megatron/training/one_logger_utils.py:329
Methodon_save_for_backward
Hook called when autograd saves a tensor for backward pass. Returns a tag to identify the tensor later.
Megatron-LM/megatron/core/pipeline_parallel/fine_grained_activation_offload.py:191
Functionon_train_start
Function will be called at the start of train function to prepare and track E2E metrics. Args: iteration (int): current iteration number
Megatron-LM/megatron/training/one_logger_utils.py:18
Functionon_training_step_end
Should be called after each training step
Megatron-LM/megatron/training/ft_integration.py:134
Functionon_training_step_start
Should be called before each training step
Megatron-LM/megatron/training/ft_integration.py:119
Methodonload
(self)
slime/slime/ray/actor_group.py:127
Methodonload_kv
(self)
slime/slime/ray/rollout.py:197
Methodonload_weights
(self)
slime/slime/ray/rollout.py:194
Functionoom_observer
(device, alloc, device_alloc, device_free)
Megatron-LM/model_provider.py:50
Functionoom_observer
(device, alloc, device_alloc, device_free)
Megatron-LM/tools/checkpoint/checkpoint_inspector.py:724
Methodoom_observer
(device, alloc, device_alloc, device_free)
slime/slime/utils/profile_utils.py:114
Functionopen_file
()
gui-rl/desktop_env/server/main.py:1286
Functionopenai_gelu
(x)
Megatron-LM/megatron/legacy/model/utils.py:52
Functionopenai_gelu
(x)
Megatron-LM/megatron/core/transformer/utils.py:61
Functionopenclaw_topk_select_loss_function
``--custom-loss-function-path`` entry point for openclaw_combine_select. Dispatches on (``args.distill_subset_mode``, ``args.hint_selection``)
openclaw-combine/openclaw_topk_select_loss.py:128
Methodoptimizer
Access underlying optimizer when only one optimizer included for backward compatibility.
Megatron-LM/megatron/core/optimizer/optimizer.py:1108
Methodorthogonalize
Orthogonalize the momentum. Args: p: The parameter tensor. i is necessary to pass param tensor in addition to momentum
Megatron-LM/megatron/core/optimizer/muon.py:109
Functionoverride_sharded_param_cpu_function
(*args, **kwargs)
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:3737
Functionoverride_sharded_param_to_function
(*args, **kwargs)
Megatron-LM/megatron/core/distributed/fsdp/src/megatron_fsdp/param_and_grad_buffer.py:3723
Methodpad
(self)
Megatron-LM/megatron/training/tokenizer/tokenizer.py:190
← previousnext →8,301–8,400 of 10,561, ranked by callers