Pretrain, finetune ANY AI model of ANY size on 1 or 10,000+ GPUs with zero code changes.
COMMITS
/ tests/models/test_tpu.py March 26, 2022
J
Update `tests/models/*.py` to use `devices` instead of `gpus` or `ipus` (#11470)
Jv Kyle Eclarin committed
March 25, 2022
K
Update tpu_cores flag with accelerator and devices flag (#12158)
Kaushik B committed
March 22, 2022
D
Remove `AcceleratorConnector.tpu_id` (#12387)
DuYicong515 committed
February 22, 2022
C
Remove `Trainer._device_type` (#11992)
Carlos Mocholí committed
December 23, 2021
A
Update strategy import statements (#11231)
Adrian Wälchli committed
December 22, 2021
A
Fix BF16 teardown for TPU precision plugin (#10990)
Adrian Wälchli committed
F
Deprecate Trainer.training_type_plugin in favor of trainer.strategy (#11141)
four4fish committed
December 21, 2021
A
Rename `TPUSpawnPlugin` to `TPUSpawnStrategy` (#11190)
Aki Nitta committed
December 20, 2021
A
Rename the `TrainingTypePlugin` base to `Strategy` (#11120)
Adrian Wälchli committed
November 25, 2021
K
Deprecate `DeviceType` in favor of `_AcceleratorType` (#10503)
Kaushik B committed
November 10, 2021
A
Fix support for dataclasses with ClassVar/InitVar in `apply_to_collection` (#9702)
a-gardner1 committed
November 8, 2021
A
Rename "master" methods to "main" in ClusterEnvironment plugins (#10103)
Adrian Wälchli committed
November 4, 2021
P
October 25, 2021
D
Mark accelerator connector as protected (#10032)
Danielle Pintz committed
October 20, 2021
K
Update strategy flag in docs (#10000)
Kaushik B committed
October 19, 2021
R
Remove deprecated `distributed_backend` from `Trainer` (#10017)
Rohit Gupta committed
October 13, 2021
K
Add `strategy` argument to Trainer (#8597)
Kaushik B committed
October 12, 2021
September 30, 2021
C
Remove unnecessary `pytest.param` usage (#9760)
Carlos Mocholí committed
September 25, 2021
D
Add `enable_progress_bar` to Trainer constructor (#9664)
Danielle Pintz committed
September 6, 2021
J
CI: precommit - docformatter (#8584)
Jirka Borovec committed
July 26, 2021
C
Replace `yapf` with `black` (#7783)
Carlos Mocholí committed
June 10, 2021
C
Clean-up after logger connector redesign 2/2 (#7631)
Carlos Mocholí committed
June 8, 2021
C
New logger connector code (#7882)
Carlos Mocholí committed
May 30, 2021
C
Some test updates (#7761)
Carlos Mocholí committed
May 27, 2021
C
Rename and move Result (#7736)
Carlos Mocholí committed
May 7, 2021
C
Use `torch.nn.utils.clip_grad_norm_` and add `clip_grad_by_value` support for TPU (#7025)
Carlos Mocholí committed
May 4, 2021
C
`TrainerState` refactor [5/5] (#7173)
Carlos Mocholí committed
April 30, 2021
K
Device updates for TPU Pod (#7243)
Kaushik B committed
April 27, 2021
K
Add `debug` flag to TPU Training Plugins (PT_XLA_DEBUG) (#7219)
Kaushik B committed
April 14, 2021
C
Fix the `gradient_clip_algorithm` has no effect issue. (#6928)
CeShine Lee committed
April 13, 2021
K
Fix sync_dist for tpus (#6950)
Kaushik B committed
April 9, 2021
K
Fix TPU Spawn gather (#6896)
Kaushik B committed
April 6, 2021
A
Add `Trainer(gradient_clip_algorithm='value'|'norm')` (#6123)
Anthony Kim committed
K
[Fix] TPU Training Type Plugin (#6816)
Kaushik B committed
March 25, 2021
K
Fix checkpoint callback & Trainer.test(_) issue for TPUs (#6654)
Kaushik B committed
March 19, 2021
K
Update Gradient Clipping for TPU Accelerator (#6576)
Kaushik B committed
March 2, 2021
T
[bugfix] TPU test hangs to barrier on 1 process (#6272)
thomas chaton committed
J
Refactor: Runif for TPU and Horovod 5/n (#6301)
Jirka Borovec committed
February 23, 2021
J
fixing miss-leading tested acc values (#5876)
Jirka Borovec committed
February 18, 2021
A
rename accelerator_backend -> accelerator (#6034)
Adrian Wälchli committed
February 17, 2021
C
[HotFix] Resolve TPU Training (#6027)
chaton committed
February 12, 2021
J
PoC: Accelerator refactor (#5743)
Justus Schock committed
February 11, 2021
R
[tests/models] refactor with BoringModel (#5507)
Rohit Gupta committed
February 8, 2021
J
Refactor simplify tests (#5861)
Jirka Borovec committed
February 6, 2021
J
formatting tests: 4/n (#5846)
Jirka Borovec committed
January 14, 2021
A
Fix pre-commit isort failure on tests/models/*.py (#5423)
Arnaud Gelas committed
January 12, 2021
J
prune check on Trainer fit result (#5453)
Jirka Borovec committed
December 21, 2020
J
fix/enable - check F401 (#5201)
Jirka Borovec committed
December 14, 2020
J
set xxx_AVAILABLE as protected (#5082)
Jirka Borovec committed