.. |
activation_checkpointing
|
08e0733e4a
Support MoE for pipeline models (#5338)
|
6 months ago |
checkpoint_engine
|
c5edc91ecb
change partititon_name to partition_name (#3700)
|
1 year ago |
comm
|
6dcced1d5c
Cleanup required_torch_version code and references. (#5370)
|
6 months ago |
compression
|
b361c72761
Update DeepSpeed copyright license to Apache 2.0 (#3111)
|
1 year ago |
data_pipeline
|
64defe65b7
Parallel map step for `DistributedDataAnalyzer` map-reduce (#5291)
|
6 months ago |
fp16
|
0896503e2f
Fix a convergence issues in TP topology caused by incorrect grad_norm. (#5411)
|
6 months ago |
pipe
|
6dcced1d5c
Cleanup required_torch_version code and references. (#5370)
|
6 months ago |
swap_tensor
|
d058d4b39b
Nvme offload checkpoint (#4707)
|
9 months ago |
zero
|
54c0687264
stage3: efficient compute of scaled_global_grad_norm (#5256)
|
6 months ago |
__init__.py
|
c56a4b9e0d
Improve universal checkpoint (#5289)
|
6 months ago |
base_optimizer.py
|
c56a4b9e0d
Improve universal checkpoint (#5289)
|
6 months ago |
bf16_optimizer.py
|
c66bc4269e
set the default to use set_to_none for clearing gradients in BF16 optimizer. (#5434)
|
6 months ago |
compiler.py
|
d274ebf347
allow debug/experimental compiler backends (#5191)
|
7 months ago |
config.py
|
ed8aed5703
fix comms dtype (#5297)
|
6 months ago |
config_utils.py
|
604d701e35
Introduce pydantic_v1 compatibility module for pydantic>=2.0.0 support (#4407)
|
1 year ago |
constants.py
|
3c0bd31288
BF16 optimizer: Improve device utilization by immediate grad update (#4975)
|
8 months ago |
dataloader.py
|
b361c72761
Update DeepSpeed copyright license to Apache 2.0 (#3111)
|
1 year ago |
eigenvalue.py
|
b361c72761
Update DeepSpeed copyright license to Apache 2.0 (#3111)
|
1 year ago |
engine.py
|
bc0f774728
Update engine.py to avoid torch warning (#5408)
|
6 months ago |
hybrid_engine.py
|
5f41bd06dd
Fix Hybrid Engine metrics printing (#4789)
|
10 months ago |
lr_schedules.py
|
ce0ebdade2
[Bug fix] WarmupCosineLR issues (#4688)
|
11 months ago |
progressive_layer_drop.py
|
b361c72761
Update DeepSpeed copyright license to Apache 2.0 (#3111)
|
1 year ago |
quantize.py
|
b361c72761
Update DeepSpeed copyright license to Apache 2.0 (#3111)
|
1 year ago |
sparse_tensor.py
|
c84c28d23b
Support cpu tensors without direct device invocation (#3842)
|
9 months ago |
state_dict_factory.py
|
b361c72761
Update DeepSpeed copyright license to Apache 2.0 (#3111)
|
1 year ago |
utils.py
|
0896503e2f
Fix a convergence issues in TP topology caused by incorrect grad_norm. (#5411)
|
6 months ago |
weight_quantizer.py
|
b361c72761
Update DeepSpeed copyright license to Apache 2.0 (#3111)
|
1 year ago |