Skip to content

[deepspeed] offload + non-cpuadam optimizer exception #22043

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
stas00 merged 3 commits into from
Mar 9, 2023
Merged

[deepspeed] offload + non-cpuadam optimizer exception #22043

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
a4a595e
[deepspeed] offload + non-cpuadam optimizer exception
stas00 Mar 9, 2023
00eed15
flip
stas00 Mar 9, 2023
e9d5ad5
revert min version
stas00 Mar 9, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions tests/deepspeed/test_deepspeed.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -426,6 +426,7 @@ def test_hf_optimizer_with_offload(self, stage, dtype):
del ds_config_dict["optimizer"] # force default HF Trainer optimizer
# force cpu offload
ds_config_dict["zero_optimization"]["offload_optimizer"]["device"] = "cpu"
ds_config_dict["zero_force_ds_cpu_optimizer"] = False # offload is not efficient w/o CPUAdam
with mockenv_context(**self.dist_env_1_gpu):
kwargs = {"local_rank": 0, "deepspeed": ds_config_dict}
kwargs[dtype] = True
Expand Down Expand Up @@ -776,6 +777,7 @@ def test_load_best_model(self, stage, dtype):
ds_config_dict = self.get_config_dict(stage)
del ds_config_dict["optimizer"] # will use HF Trainer optimizer
del ds_config_dict["scheduler"] # will use HF Trainer scheduler
ds_config_dict["zero_force_ds_cpu_optimizer"] = False # offload is not efficient w/o CPUAdam
# must use this setting to get the reload path exercised
ds_config_dict["zero_optimization"]["stage3_gather_16bit_weights_on_model_save"] = True

Expand Down