ros/LLaMA-Factory
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

drafting ray integration

Browse Source 
Signed-off-by: Kourosh Hakhamaneshi <kourosh@anyscale.com>

Former-commit-id: 19c12ddae9350f6e25a270fe3372f5b9094cf960
This commit is contained in:
Kourosh Hakhamaneshi
2024-12-30 16:48:52 -08:00
committed by hiyouga
parent 5ccc607222
commit 8683582300
9 changed files with 143 additions and 21 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
src/llamafactory/cli.py
View File

@@ -27,7 +27,7 @@ from .extras.env import VERSION, print_env
from .extras.misc import get_device_count
from .train.tuner import export_model, run_exp
from .webui.interface import run_web_demo, run_web_ui
from .integrations.ray.ray_utils import should_use_ray
USAGE = (
"-" * 70
@@ -87,7 +87,8 @@ def main():
export_model()
elif command == Command.TRAIN:
force_torchrun = os.getenv("FORCE_TORCHRUN", "0").lower() in ["true", "1"]
if force_torchrun or get_device_count() > 1:
use_ray = should_use_ray()
if force_torchrun or (get_device_count() > 1 and not use_ray):
master_addr = os.getenv("MASTER_ADDR", "127.0.0.1")
master_port = os.getenv("MASTER_PORT", str(random.randint(20001, 29999)))
logger.info_rank0(f"Initializing distributed tasks at: {master_addr}:{master_port}")