SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

fix the deepspeed tests (#26021)

Browse Source 
* fix the deepspeed tests

* resolve comment
This commit is contained in:
Sourab Mangrulkar
2023-09-13 10:26:53 +05:30
committed by GitHub
parent 73b13ac099
commit b477327394
1 changed files with 5 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
tests/deepspeed/test_deepspeed.py
View File

@@ -379,19 +379,16 @@ class TrainerIntegrationDeepSpeed(TrainerIntegrationDeepSpeedWithCustomConfig, T
self.assertNotEqual(new_a, a) self.assertNotEqual(new_a, a)
def test_hf_scheduler_ds_optimizer(self): def test_hf_scheduler_ds_optimizer(self):
a = 0
with mockenv_context(**self.dist_env_1_gpu): with mockenv_context(**self.dist_env_1_gpu):
ds_config_zero2_dict = self.get_config_dict(ZERO2) ds_config_zero2_dict = self.get_config_dict(ZERO2)
del ds_config_zero2_dict["scheduler"] # force default HF Trainer scheduler del ds_config_zero2_dict["scheduler"] # force default HF Trainer scheduler
ds_config_zero2_dict["zero_optimization"]["offload_optimizer"]["device"] = "none" ds_config_zero2_dict["zero_optimization"]["offload_optimizer"]["device"] = "none"
ds_config_zero2_dict["fp16"]["initial_scale_power"] = 1 # force optimizer on the first step ds_config_zero2_dict["fp16"]["initial_scale_power"] = 1 # force optimizer on the first step
trainer = get_regression_trainer(local_rank=0, fp16=True, deepspeed=ds_config_zero2_dict) trainer = get_regression_trainer(a=a, local_rank=0, fp16=True, deepspeed=ds_config_zero2_dict)
with self.assertRaises(Exception) as context: trainer.train()
trainer.train() new_a = trainer.model.a.item()
self.assertIn( self.assertNotEqual(new_a, a)
"Found `optimizer` configured in the DeepSpeed config, but no `scheduler`. "
"Please configure a scheduler in the DeepSpeed config.",
str(context.exception),
)
@require_deepspeed_aio @require_deepspeed_aio
def test_stage3_nvme_offload(self): def test_stage3_nvme_offload(self):