github/ColossalAI
1
0
Fork 0
mirror of https://github.com/hpcaitech/ColossalAI.git synced 2025-09-05 02:51:59 +00:00
Code Issues Packages Projects Releases Wiki Activity

[misc] update pre-commit and run all files (#4752)

Browse Source 
* [misc] update pre-commit

* [misc] run pre-commit

* [misc] remove useless configuration files

* [misc] ignore cuda for clang-format
This commit is contained in:
Hongxin Liu
2023-09-19 14:20:26 +08:00
committed by GitHub
parent 3c6b831c26
commit 079bf3cb26
1268 changed files with 50037 additions and 38444 deletions
Download Patch File Download Diff File Expand all files Collapse all files

82
tests/test_booster/test_plugin/test_gemini_plugin.py
View File

@@ -17,7 +17,7 @@ from tests.kit.model_zoo import model_zoo
def run_fn(init_method, model_fn, data_gen_fn, output_transform_fn) -> Optional[str]:
try:
if init_method == 'lazy':
if init_method == "lazy":
ctx = LazyInitContext()
else:
ctx = nullcontext()
@@ -30,13 +30,13 @@ def run_fn(init_method, model_fn, data_gen_fn, output_transform_fn) -> Optional[
data = data_gen_fn()
data = {
k: v.to('cuda') if torch.is_tensor(v) or 'Tensor' in v.__class__.__name__ else v for k, v in data.items()
k: v.to("cuda") if torch.is_tensor(v) or "Tensor" in v.__class__.__name__ else v for k, v in data.items()
}
model, optimizer, criterion, _, _ = booster.boost(model, optimizer, criterion)
for n, p in model.named_parameters():
assert isinstance(p, ColoParameter), f'{n} is not a ColoParameter'
assert isinstance(p, ColoParameter), f"{n} is not a ColoParameter"
output = model(**data)
output = output_transform_fn(output)
@@ -55,47 +55,65 @@ def run_fn(init_method, model_fn, data_gen_fn, output_transform_fn) -> Optional[
# @parameterize('init_method', ['lazy', 'none', 'colo'])
@parameterize('subset', ['torchvision', 'transformers', 'diffusers'])
@parameterize('init_method', ['none'])
def check_gemini_plugin(subset: str, init_method: str = 'none', early_stop: bool = True):
@parameterize("subset", ["torchvision", "transformers", "diffusers"])
@parameterize("init_method", ["none"])
def check_gemini_plugin(subset: str, init_method: str = "none", early_stop: bool = True):
"""check gemini plugin over model zoo
Args:
early_stop (bool, optional): Whether to stop when getting the first error. Defaults to True.
"""
is_support_meta = is_compatible_with_meta()
if not is_support_meta and init_method == 'lazy':
if not is_support_meta and init_method == "lazy":
return
passed_models = []
failed_info = {} # (model_name, error) pair
failed_info = {} # (model_name, error) pair
for name, (model_fn, data_gen_fn, output_transform_fn, _, _) in model_zoo.get_sub_registry(subset).items():
# These models lead to CUDA error
if name in ('diffusers_auto_encoder_kl', 'diffusers_vq_model', 'diffusers_unet2d_model', 'timm_resmlp',
'timm_gmixer_12_224', 'timm_gmlp_b16_224', 'timm_mixer_b16_224', 'timm_convnext',
'torchvision_convnext_base'):
if name in (
"diffusers_auto_encoder_kl",
"diffusers_vq_model",
"diffusers_unet2d_model",
"timm_resmlp",
"timm_gmixer_12_224",
"timm_gmlp_b16_224",
"timm_mixer_b16_224",
"timm_convnext",
"torchvision_convnext_base",
):
continue
# These models are not compatible with gemini
if name in [
'timm_convit',
'timm_dm_nfnet',
'torchvision_vit_b_16',
'transformers_t5',
'transformers_t5_for_conditional_generation',
'transformers_t5_encoder_model', # does not support apex rmsnorm
'transformers_chatglm',
'transformers_sam',
'transformers_vit',
'transformers_gpt_double_heads', # TODO check why does the model fail to run using Gemini
"timm_convit",
"timm_dm_nfnet",
"torchvision_vit_b_16",
"transformers_t5",
"transformers_t5_for_conditional_generation",
"transformers_t5_encoder_model", # does not support apex rmsnorm
"transformers_chatglm",
"transformers_sam",
"transformers_vit",
"transformers_gpt_double_heads", # TODO check why does the model fail to run using Gemini
]:
continue
if init_method == 'lazy' and name in [
'timm_convmixer', 'timm_vision_transformer', 'timm_deit', 'timm_deit3', 'timm_inception_v3',
'timm_tnt_b_patch16_224', 'timm_rexnet', 'torchvision_densenet121', 'torchvision_efficientnet_b0',
'torchvision_mobilenet_v2', 'torchvision_mnasnet0_5', 'torchvision_regnet_x_16gf',
'torchvision_shufflenet_v2_x0_5', 'torchvision_efficientnet_v2_s'
if init_method == "lazy" and name in [
"timm_convmixer",
"timm_vision_transformer",
"timm_deit",
"timm_deit3",
"timm_inception_v3",
"timm_tnt_b_patch16_224",
"timm_rexnet",
"torchvision_densenet121",
"torchvision_efficientnet_b0",
"torchvision_mobilenet_v2",
"torchvision_mnasnet0_5",
"torchvision_regnet_x_16gf",
"torchvision_shufflenet_v2_x0_5",
"torchvision_efficientnet_v2_s",
]:
continue
err = run_fn(init_method, model_fn, data_gen_fn, output_transform_fn)
@@ -108,15 +126,15 @@ def check_gemini_plugin(subset: str, init_method: str = 'none', early_stop: bool
break
if dist.get_rank() == 0:
print(f'Init method: {init_method}')
print(f'Passed models({len(passed_models)}): {passed_models}\n\n')
print(f'Failed models({len(failed_info)}): {list(failed_info.keys())}\n\n')
assert len(failed_info) == 0, '\n'.join([f'{k}: {v}' for k, v in failed_info.items()])
print(f"Init method: {init_method}")
print(f"Passed models({len(passed_models)}): {passed_models}\n\n")
print(f"Failed models({len(failed_info)}): {list(failed_info.keys())}\n\n")
assert len(failed_info) == 0, "\n".join([f"{k}: {v}" for k, v in failed_info.items()])
def run_dist(rank, world_size, port, early_stop: bool = True):
# init dist env
colossalai.launch(config=dict(), rank=rank, world_size=world_size, port=port, host='localhost')
colossalai.launch(config=dict(), rank=rank, world_size=world_size, port=port, host="localhost")
check_gemini_plugin(early_stop=early_stop)
@@ -125,5 +143,5 @@ def test_gemini_plugin(early_stop: bool = True):
spawn(run_dist, 4, early_stop=early_stop)
if __name__ == '__main__':
if __name__ == "__main__":
test_gemini_plugin(early_stop=False)