github/ColossalAI
1
0
Fork 0
mirror of https://github.com/hpcaitech/ColossalAI.git synced 2025-08-06 02:24:31 +00:00
Code Issues Packages Projects Releases Wiki Activity

add uuid to rollout log

Browse Source
This commit is contained in:
YeAnbang 2025-05-20 09:45:56 +08:00
parent 107470a360
commit f8bd2db33f
2 changed files with 6 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
applications/ColossalChat/coati/distributed/launch.py
View File

@ -56,7 +56,7 @@ def launch_distributed(
eval_save_dir: Optional[str] = None,
eval_generation_config: Optional[Dict[str, Any]] = None,
log_rollout_interval: int = 20,
rollout_log_file: str = "./rollout_log.jsonl",
rollout_save_dir: str = "./rollout",
):
if core_algo not in ALGO_MAP:
raise NotImplementedError(f"{core_algo} is not supported yet.")
@ -74,6 +74,10 @@ def launch_distributed(
run_name = f"{inference_backend}_bs_{train_batch_size * train_dp_size}_temp_{generate_config['temperature']:.01f}_top_p_{generate_config['top_p']:.02f}"
wandb_group_name = str(uuid.uuid4())
rollout_log_file = os.path.join(
rollout_save_dir,
f"{project_name.replace(' ','_')}_run_{wandb_group_name}.jsonl",
)
procs = []
for i in range(num_producers):

2
applications/ColossalChat/rl_example.py
View File

@ -273,5 +273,5 @@ if __name__ == "__main__":
eval_save_dir=os.path.join(args.eval_save_dir, args.project.replace(" ", "_")),
eval_generation_config=eval_generation_config,
log_rollout_interval=20,
rollout_log_file=os.path.join(args.rollout_save_dir, args.project.replace(" ", "_") + ".jsonl"),
rollout_save_dir=args.rollout_save_dir,
)