p12475398
/
LLaMA-Factory-Mirror
forked from p04798526/LLaMA-Factory-Mirror
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Update cli.py

This commit is contained in:
hoshi-hiyouga 2024-06-06 23:38:09 +08:00 committed by GitHub
parent 90ed3cae92
commit 751dd77bc0
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 0 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
src/llamafactory/cli.py
View File

@ -71,10 +71,6 @@ def main():
export_model() export_model()
elif command == Command.TRAIN: elif command == Command.TRAIN:
if get_device_count() > 0: if get_device_count() > 0:
# NOTE (MengqingCao): why use torchrun when only one accelerator is available?
# DeepSpeed only warp model with DeepSpeedEngine when launching by distributed launcher,
# e.g., torchrun, causing some feature missing
# sa: https://github.com/huggingface/transformers/issues/24309
master_addr = os.environ.get("MASTER_ADDR", "127.0.0.1") master_addr = os.environ.get("MASTER_ADDR", "127.0.0.1")
master_port = os.environ.get("MASTER_PORT", str(random.randint(20001, 29999))) master_port = os.environ.get("MASTER_PORT", str(random.randint(20001, 29999)))
logger.info("Initializing distributed tasks at: {}:{}".format(master_addr, master_port)) logger.info("Initializing distributed tasks at: {}:{}".format(master_addr, master_port))