| .. |
|
launchers
|
Fix `is_interactive_compatible` logic after AcceleratorConnector rewrite (#12008)
|
2022-02-22 19:20:54 +05:30 |
|
__init__.py
|
Add process launchers (#11643)
|
2022-02-17 21:16:51 +00:00 |
|
bagua.py
|
Rewrite accelerator_connector (#11448)
|
2022-02-17 23:38:39 +00:00 |
|
ddp.py
|
Add `SyncBatchNormPlugin` (#11754)
|
2022-03-01 19:41:40 +05:30 |
|
ddp2.py
|
Rewrite accelerator_connector (#11448)
|
2022-02-17 23:38:39 +00:00 |
|
ddp_spawn.py
|
Add `SyncBatchNormPlugin` (#11754)
|
2022-03-01 19:41:40 +05:30 |
|
deepspeed.py
|
Refactor `Strategy._move_optimizer_states` as utility functions (#11758)
|
2022-02-18 08:36:07 +00:00 |
|
dp.py
|
Rewrite accelerator_connector (#11448)
|
2022-02-17 23:38:39 +00:00 |
|
fully_sharded.py
|
Add `SyncBatchNormPlugin` (#11754)
|
2022-03-01 19:41:40 +05:30 |
|
horovod.py
|
Support gradient accumulation using Horovod's `backward_passes_per_step` (#11911)
|
2022-02-19 02:54:04 +01:00 |
|
ipu.py
|
[IPU] Support manually instantiating the `poptorch.DataLoader` (#12116)
|
2022-02-28 09:36:26 +00:00 |
|
parallel.py
|
Add `SyncBatchNormPlugin` (#11754)
|
2022-03-01 19:41:40 +05:30 |
|
sharded.py
|
Rewrite accelerator_connector (#11448)
|
2022-02-17 23:38:39 +00:00 |
|
sharded_spawn.py
|
Rewrite accelerator_connector (#11448)
|
2022-02-17 23:38:39 +00:00 |
|
single_device.py
|
Rewrite accelerator_connector (#11448)
|
2022-02-17 23:38:39 +00:00 |
|
single_tpu.py
|
Rewrite accelerator_connector (#11448)
|
2022-02-17 23:38:39 +00:00 |
|
strategy.py
|
Refactor `Strategy._move_optimizer_states` as utility functions (#11758)
|
2022-02-18 08:36:07 +00:00 |
|
strategy_registry.py
|
Rewrite accelerator_connector (#11448)
|
2022-02-17 23:38:39 +00:00 |
|
tpu_spawn.py
|
Remove `DDPSpawnStrategy.get_mp_spawn_kwargs` in favor of launchers (#11966)
|
2022-02-22 11:28:21 +00:00 |
|
utils.py
|
Introduce strategies directory for Training Strategies (#11226)
|
2021-12-22 20:23:30 +00:00 |
d4d197070f