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

[deepspeed] test on one node 2 gpus max (#11237)

Browse Source 
* test on one node 2 gpus max

* fix the other place

* refactor

* fix

* cleanup

* more exact version
This commit is contained in:
Stas Bekman
2021-04-14 11:06:59 -07:00
committed by GitHub
parent 25e1af36e0
commit 83206ca6a8
2 changed files with 11 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
setup.py
View File

@@ -90,7 +90,7 @@ _deps = [
"cookiecutter==1.7.2", "cookiecutter==1.7.2",
"dataclasses", "dataclasses",
"datasets", "datasets",
"deepspeed>0.3.13", "deepspeed>=0.3.14",
"docutils==0.16.0", "docutils==0.16.0",
"fairscale>0.3", "fairscale>0.3",
"faiss-cpu", "faiss-cpu",

15
tests/deepspeed/test_deepspeed.py
View File

@@ -595,8 +595,7 @@ class TestDeepSpeedWithLauncher(TestCasePlus):
ds_args = f"--deepspeed {self.test_file_dir_str}/ds_config_{stage}.json".split() ds_args = f"--deepspeed {self.test_file_dir_str}/ds_config_{stage}.json".split()
script = [f"{self.examples_dir_str}/seq2seq/run_translation.py"] script = [f"{self.examples_dir_str}/seq2seq/run_translation.py"]
num_gpus = get_gpu_count() if distributed else 1 launcher = self.get_launcher(distributed)
launcher = f"deepspeed --num_gpus {num_gpus}".split()
cmd = launcher + script + args + ds_args cmd = launcher + script + args + ds_args
# keep for quick debug # keep for quick debug
@@ -629,11 +628,9 @@ class TestDeepSpeedWithLauncher(TestCasePlus):
--block_size 128 --block_size 128
""".split() """.split()
distributed = True
ds_args = f"--deepspeed {self.test_file_dir_str}/ds_config_{stage}.json".split() ds_args = f"--deepspeed {self.test_file_dir_str}/ds_config_{stage}.json".split()
script = [f"{self.examples_dir_str}/language-modeling/run_clm.py"] script = [f"{self.examples_dir_str}/language-modeling/run_clm.py"]
num_gpus = get_gpu_count() if distributed else 1 launcher = self.get_launcher(distributed=True)
launcher = f"deepspeed --num_gpus {num_gpus}".split()
cmd = launcher + script + args + ds_args cmd = launcher + script + args + ds_args
# keep for quick debug # keep for quick debug
@@ -641,3 +638,11 @@ class TestDeepSpeedWithLauncher(TestCasePlus):
execute_subprocess_async(cmd, env=self.get_env()) execute_subprocess_async(cmd, env=self.get_env())
return output_dir return output_dir
def get_launcher(self, distributed=False):
# 1. explicitly set --num_nodes=1 just in case these tests end up run on a multi-node setup
# - it won't be able to handle that
# 2. for now testing with just 2 gpus max (since some quality tests may give different
# results with mode gpus because we use very little data)
num_gpus = min(2, get_gpu_count()) if distributed else 1
return f"deepspeed --num_nodes 1 --num_gpus {num_gpus}".split()