MCPcopy Create free account

hub / github.com/AIS-SNU/Smart-Infinity / functions

Functions4,989 in github.com/AIS-SNU/Smart-Infinity

↓ 590 callersMethodappend
(self, raw_name, record_name, latency, msg_size)
deepspeed/utils/comms_logging.py:98
↓ 447 callersMethodsize
(self, index)
deepspeed/runtime/data_pipeline/data_sampling/indexed_dataset.py:206
↓ 334 callersFunctionget_accelerator
()
deepspeed/accelerator/real_accelerator.py:38
↓ 277 callersMethoddata_ptr
deepspeed/ops/csrc/aio/py_lib/deepspeed_aio_thread.cpp:31
↓ 272 callersMethodnumel
(self)
deepspeed/runtime/swap_tensor/optimizer_utils.py:59
↓ 241 callersMethodappend
(self, other)
DeepSpeedExample/megatron/deprecated_data_utils/tokenization.py:109
↓ 156 callersFunctionget_scalar_param
(param_dict, param_name, param_default_value)
deepspeed/runtime/config_utils.py:186
↓ 148 callersMethodget_rank
(self, group=None)
deepspeed/comm/torch.py:221
↓ 138 callersMethodget_rank
Return the global rank of a process via its coordinates. Coordinates are specified as kwargs. For example: >>> X = ProcessTopolo
deepspeed/runtime/pipe/topology.py:49
↓ 138 callersMethodget_world_size
(self, group=None)
deepspeed/comm/torch.py:224
↓ 129 callersFunctionprint_rank_0
If distributed is initialized print only on rank 0.
DeepSpeedExample/megatron/__init__.py:35
↓ 118 callersMethodreshape
(self, target_3d_desc, verbose=False)
deepspeed/checkpoint/reshape_3d_utils.py:24
↓ 118 callersMethodsize
(self, index)
DeepSpeedExample/megatron/data/indexed_dataset.py:197
↓ 116 callersMethodload
(self, verbose=True)
deepspeed/ops/op_builder/builder.py:443
↓ 94 callersFunctionget_args
Return arguments.
DeepSpeedExample/megatron/global_vars.py:34
↓ 93 callersMethodget
deepspeed/ops/csrc/includes/memory_access_utils.h:74
↓ 90 callersMethodextend
(self, other)
DeepSpeedExample/megatron/deprecated_data_utils/tokenization.py:122
↓ 90 callersFunctionsee_memory_usage
(message, force=False)
deepspeed/runtime/utils.py:775
↓ 82 callersMethodcurrent_device_name
(self)
deepspeed/accelerator/cpu_accelerator.py:37
↓ 81 callersMethodall_reduce
(self, tensor, op=ReduceOp.SUM, group=None, async_op=False)
deepspeed/comm/ccl.py:53
↓ 74 callersMethodwarning
(self, msg)
deepspeed/ops/op_builder/builder.py:425
↓ 70 callersMethodread
deepspeed/ops/csrc/aio/py_lib/deepspeed_py_aio_handle.cpp:63
↓ 62 callersMethodis_initialized
(self)
deepspeed/comm/ccl.py:47
↓ 57 callersMethodcopy
(self, dst, src, int8=False, allocate_tensor=False)
deepspeed/module_inject/replace_module.py:93
↓ 55 callersMethodbarrier
(self, group=None, async_op=False)
deepspeed/comm/ccl.py:61
↓ 52 callersMethoddtype
(self)
deepspeed/runtime/swap_tensor/optimizer_utils.py:85
↓ 50 callersMethodget
Retrieves a single item from the dataset with the option to only return a portion of the item. get(idx) is the same as [idx] but get
deepspeed/runtime/data_pipeline/data_sampling/indexed_dataset.py:532
↓ 50 callersFunctionlog_dist
(message, ranks=None, level=logging.INFO)
deepspeed/utils/logging.py:75
↓ 48 callersMethodGetCurrentStream
deepspeed/ops/csrc/transformer/inference/includes/inference_context.h:211
↓ 48 callersFunctionwrapFunc
(func, funcFlopCompute)
deepspeed/profiling/flops_profiler/profiler.py:741
↓ 43 callersMethodupdate
Update the tuner with what configurations have been evaluated and their performance results
deepspeed/autotuning/tuner/base_tuner.py:35
↓ 41 callersMethodLongTensor
(self)
deepspeed/accelerator/cpu_accelerator.py:213
↓ 41 callersMethoddevice
(self, device_index=None)
deepspeed/accelerator/cpu_accelerator.py:28
↓ 40 callersMethodstart
(self)
deepspeed/utils/timer.py:191
↓ 39 callersMethodsynchronize
(self, device_index=None)
deepspeed/accelerator/cpu_accelerator.py:56
↓ 38 callersMethodapply
(self, module, enable_cuda_graph=True)
deepspeed/module_inject/containers/vae.py:30
↓ 38 callersMethodnew_group
(self)
deepspeed/comm/backend.py:40
↓ 36 callersMethoddevice_name
(self, device_index=None)
deepspeed/accelerator/cpu_accelerator.py:25
↓ 36 callersMethodget_model_parallel_group
(self)
deepspeed/runtime/pipe/topology.py:442
↓ 36 callersMethodtype
()
deepspeed/runtime/sparse_tensor.py:38
↓ 34 callersMethodexists
(path)
deepspeed/runtime/data_pipeline/data_sampling/indexed_dataset.py:210
↓ 34 callersMethodget_model_parallel_rank
(self)
deepspeed/runtime/pipe/topology.py:436
↓ 34 callersMethodget_rng_state
(self, device_index=None)
deepspeed/accelerator/cpu_accelerator.py:68
↓ 33 callersMethodstop
(self, global_step=False, report_speed=True)
deepspeed/utils/timer.py:198
↓ 32 callersMethodtranspose
(self)
deepspeed/module_inject/containers/base.py:264
↓ 31 callersMethodstart
Start the timer.
DeepSpeedExample/megatron/global_vars.py:164
↓ 30 callersMethodget_data_parallel_group
The group of ranks within the same stage of all pipelines.
deepspeed/runtime/pipe/topology.py:430
↓ 29 callersMethodGetCublasHandle
deepspeed/ops/csrc/transformer/inference/includes/inference_context.h:233
↓ 29 callersMethodapply
(self, args)
DeepSpeedExample/megatron/deprecated_data_utils/configure_data.py:31
↓ 29 callersMethodget
Retrieves a single item from the dataset with the option to only return a portion of the item. get(idx) is the same as [idx] but get
DeepSpeedExample/megatron/data/indexed_dataset.py:500
↓ 29 callersMethodget_data_parallel_rank
Which pipeline this rank resides in.
deepspeed/runtime/pipe/topology.py:422
↓ 28 callersMethodbroadcast
(self, tensor, src, group=None, async_op=False)
deepspeed/comm/ccl.py:50
↓ 28 callersMethodmakedirs
(self, path, exist_ok=False)
deepspeed/runtime/checkpoint_engine/checkpoint_engine.py:19
↓ 28 callersMethodstop
Stop the timer.
DeepSpeedExample/megatron/global_vars.py:171
↓ 27 callersMethodcurrent_stream
(self, device_index=None)
deepspeed/accelerator/cpu_accelerator.py:92
↓ 27 callersMethodwall_clock_breakdown
(self)
deepspeed/runtime/engine.py:568
↓ 26 callersMethodsend
(self, tensor, dst, group=None, tag=0)
deepspeed/comm/torch.py:185
↓ 25 callersMethodcurrent_device
(self)
deepspeed/accelerator/cpu_accelerator.py:34
↓ 25 callersMethodwait
deepspeed/ops/csrc/aio/py_lib/deepspeed_py_aio_handle.cpp:174
↓ 24 callersMethodds_summary
(slf: torch.Tensor, use_debug_name: bool = False)
deepspeed/runtime/zero/partition_parameters.py:988
↓ 23 callersMethodbfloat16_enabled
(self)
deepspeed/runtime/engine.py:759
↓ 23 callersMethodwrite
Write timers to a tensorboard writer
DeepSpeedExample/megatron/global_vars.py:211
↓ 22 callersMethodFloatTensor
(self)
deepspeed/accelerator/cpu_accelerator.py:201
↓ 22 callersMethodall_gather
(param_list=None, async_op=False, hierarchy=0)
deepspeed/runtime/zero/partition_parameters.py:874
↓ 22 callersMethodallocate_tensor
(self, size)
deepspeed/runtime/zero/contiguous_memory_allocator.py:51
↓ 22 callersMethodis_rocm_pytorch
()
deepspeed/ops/op_builder/builder.py:164
↓ 21 callersMethodget_model_parallel_world_size
(self)
deepspeed/runtime/pipe/topology.py:439
↓ 21 callersFunctionmaybe_copy
(module, sd, weight_quantizer, mp_replace, dst_nam
deepspeed/module_inject/policy.py:140
↓ 21 callersFunctionprint_rank_0
(message, debug=False, force=False)
deepspeed/runtime/zero/stage3.py:33
↓ 21 callersMethodsizes
(self)
deepspeed/runtime/data_pipeline/data_sampling/indexed_dataset.py:546
↓ 21 callersMethodstrip_empty_entries
Drop any empty strings from the list of compile and link flags
deepspeed/ops/op_builder/builder.py:321
↓ 20 callersFunctionget_caller_func
(frame=3)
deepspeed/utils/comms_logging.py:10
↓ 20 callersMethodoptimizer_name
(self)
deepspeed/runtime/engine.py:638
↓ 19 callersFunctionget_tensor_model_parallel_group
Get the tensor model parallel group the caller rank belongs to.
DeepSpeedExample/megatron/core/parallel_state.py:348
↓ 19 callersFunctionprint_rank_0
(message, debug=False, force=False)
deepspeed/runtime/zero/partition_parameters.py:85
↓ 19 callersMethodrecv
(self, tensor, src=None, group=None, tag=0)
deepspeed/comm/torch.py:188
↓ 18 callersMethodadd
Track the rng state.
deepspeed/runtime/activation_checkpointing/checkpointing.py:151
↓ 18 callersMethodelapsed
Calculate the elapsed time.
deepspeed/utils/timer.py:90
↓ 18 callersFunctionget_model_chunk_id
Helper method to get the model chunk ID given the iteration number.
DeepSpeedExample/megatron/core/pipeline_parallel/schedules.py:487
↓ 18 callersMethodsave
(self, state_dict, path: str)
deepspeed/runtime/checkpoint_engine/checkpoint_engine.py:22
↓ 17 callersMethod_start_timer
(self, name)
deepspeed/runtime/swap_tensor/optimizer_utils.py:668
↓ 17 callersMethod_stop_timer
(self, name)
deepspeed/runtime/swap_tensor/optimizer_utils.py:672
↓ 17 callersFunctionconvert_to_shared
deepspeed/ops/csrc/includes/memory_access_utils.h:586
↓ 17 callersMethodexists
check if the filepath for a text tokenizer exists
DeepSpeedExample/megatron/deprecated_data_utils/tokenization.py:437
↓ 17 callersMethodfp16_enabled
(self)
deepspeed/runtime/engine.py:756
↓ 17 callersMethodmanual_seed
(self, seed)
deepspeed/accelerator/cpu_accelerator.py:71
↓ 17 callersMethodzero_optimization
(self)
deepspeed/runtime/engine.py:668
↓ 16 callersMethoddevice
(self)
deepspeed/runtime/swap_tensor/optimizer_utils.py:82
↓ 16 callersMethodempty_cache
(self)
deepspeed/accelerator/cpu_accelerator.py:103
↓ 16 callersFunctionget_tensor_model_parallel_world_size
Return world size for the tensor model parallel group.
DeepSpeedExample/megatron/core/parallel_state.py:437
↓ 16 callersFunctionget_tokenizer
Return tokenizer.
DeepSpeedExample/megatron/global_vars.py:40
↓ 16 callersMethodstream
(self, stream)
deepspeed/accelerator/cpu_accelerator.py:88
↓ 15 callersFunction_prod
(dims)
deepspeed/profiling/flops_profiler/profiler.py:437
↓ 15 callersMethod_valid_micro_batch
(self, micro_batch_id)
deepspeed/runtime/pipe/schedule.py:80
↓ 15 callersMethoddump
deepspeed/ops/csrc/aio/common/deepspeed_aio_types.cpp:41
↓ 15 callersMethodget_data_parallel_world_size
The number of pipelines.
deepspeed/runtime/pipe/topology.py:426
↓ 15 callersMethodload
(self, mp_world_size, mp_rank, module_key=AUTO_MODULE_KEY,
deepspeed/runtime/state_dict_factory.py:57
↓ 15 callersMethodlog
(self, data, step=None, commit=None, sync=None)
deepspeed/monitor/wandb.py:27
↓ 15 callersMethodmean
(self)
deepspeed/utils/timer.py:106
↓ 15 callersMethodpartition_numel
()
deepspeed/runtime/zero/partition_parameters.py:981
next →1–100 of 4,989, ranked by callers