Pretrain, finetune ANY AI model of ANY size on 1 or 10,000+ GPUs with zero code changes.
COMMITS
/ docs/source-pytorch/advanced/model_parallel.rst January 9, 2024
J
docs: update references to ext. integrations (#19248)
Jirka Borovec committed
August 25, 2023
A
Update titles in FSDP docs (#18389)
Adrian Wälchli committed
August 16, 2023
A
Revamp model parallel docs (2/n) (#18325)
Adrian Wälchli committed
A
Revamp model parallel docs (1/n) (#18314)
Adrian Wälchli committed
July 19, 2023
C
Relax the assumption that the root module is FSDP wrapped (#18054)
Carlos Mocholí committed
July 17, 2023
A
Update deepspeed model-parallel docs (#18091)
Adrian Wälchli committed
July 15, 2023
C
Support sets for policies in FSDP (#18084)
Carlos Mocholí committed
July 14, 2023
C
Support custom policies for activation checkpointing with FSDP (#18045)
Carlos Mocholí committed
C
Add `Trainer.init_module` and `LightningModule.configure_model` (#18004)
Carlos Mocholí committed
July 10, 2023
C
Don't reapply activation checkpointing (#18006)
Carlos Mocholí committed
June 21, 2023
J
docs: include external pages (#17826)
Jirka Borovec committed
June 8, 2023
O
Custom DeepSpeed config format error with offload_optimizer (#17790)
one-matrix committed
April 11, 2023
A
Simplified setup of optimizers in FSDP (#17309)
Adrian Wälchli committed
March 21, 2023
B
Support BaguaStrategy with external implementation (#17029)
bafangchaoshi committed
March 13, 2023
J
docs: adding hivemind (#17038)
Jirka Borovec committed
March 9, 2023
C
Review APIs experimental status (#17012)
Carlos Mocholí committed
February 27, 2023
J
docs: update `pytorch_lightning` imports (#16864)
Jirka Borovec committed
February 16, 2023
A
Update Colossal AI docs and integration (#16778)
Adrian Wälchli committed
February 6, 2023
A
Set `find_unused_parameters=False` as the default (#16611)
Adrian Wälchli committed
February 5, 2023
S
Move fsdp_native to fine-tuning recommendation (#16630)
Sebastian Raschka committed
January 30, 2023
A
Minor formatting fix on model_parallel docs (#16565)
Andrea Tupini committed
January 25, 2023
C
Remove the "native" suffix from the codebase (#16490)
Carlos Mocholí committed
January 17, 2023
C
Deprecate the FairScale integration (#16353)
Carlos Mocholí committed
January 10, 2023
N
Fixes in sphinx docs links (#16255)
Nikhil Shenoy committed
December 7, 2022
A
Simplify enabling CPU offload in FSDP (#15832)
Adrian Wälchli committed
December 6, 2022
A
Activation checkpointing in FSDP without boilerplate (#15826)
Adrian Wälchli committed
December 1, 2022
C
Remove checks for torch greater than 1.10 (#15846)
Carlos Mocholí committed
November 10, 2022
C
Drop PyTorch 1.9 support (#15347)
Carlos Mocholí committed
November 8, 2022
October 13, 2022
H
[docs] Docs for ColossalaiStrategy (#15093)
HELSON committed
September 5, 2022
R
Add auto wrapping support for `DDPFullyShardedStrategy` (#14383)
Rohit Gupta committed
August 26, 2022
R
Add auto wrapping for `DDPFullyShardedNativeStrategy` (#14252)
Rohit Gupta committed
August 12, 2022
R
Add docs for `fsdp_native` (#14108)
Rohit Gupta committed
June 21, 2022
R
Update old PL links (#13349)
Rohit Gupta committed
June 15, 2022
J
Future 3/n: docs adjustment (#13299)
Jirka Borovec committed