chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 11 Packages Projects Releases Wiki Activity

Fix duplicate download log messages in multi-process environment (#14299)

Browse Source
This commit is contained in:
alisonshao
2025-12-02 09:33:18 -08:00
committed by GitHub
parent df1f31241b
commit 25a6be4930
1 changed files with 22 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files

49
python/sglang/srt/model_loader/weight_utils.py
View File

@@ -322,12 +322,10 @@ def find_local_hf_snapshot_dir(
incomplete_files = glob.glob(os.path.join(blobs_dir, "*.incomplete"))
if incomplete_files:
logger.info(
"Found %d .incomplete files in %s for %s. "
"Will clean up and re-download.",
len(incomplete_files),
blobs_dir,
model_name_or_path,
log_info_on_rank0(
logger,
f"Found {len(incomplete_files)} .incomplete files in {blobs_dir} for "
f"{model_name_or_path}. Will clean up and re-download.",
)
_cleanup_corrupted_model_cache(
model_name_or_path,
@@ -367,22 +365,20 @@ def find_local_hf_snapshot_dir(
if not is_valid:
if corrupted_files:
# Selective cleanup: only remove corrupted files
logger.info(
"Found %d corrupted file(s) for %s: %s. "
log_info_on_rank0(
logger,
f"Found {len(corrupted_files)} corrupted file(s) for "
f"{model_name_or_path}: {error_msg}. "
"Will selectively clean and re-download only these files.",
len(corrupted_files),
model_name_or_path,
error_msg,
)
_cleanup_corrupted_files_selective(model_name_or_path, corrupted_files)
return None
else:
# Cannot selectively clean (e.g., missing shards) - remove entire cache
logger.info(
"Validation failed for %s: %s. "
log_info_on_rank0(
logger,
f"Validation failed for {model_name_or_path}: {error_msg}. "
"Will remove entire cache and re-download.",
model_name_or_path,
error_msg,
)
_cleanup_corrupted_model_cache(
model_name_or_path, found_local_snapshot_dir, error_msg
@@ -400,28 +396,27 @@ def find_local_hf_snapshot_dir(
"adapter_model.safetensors",
]:
if not _validate_safetensors_file(f):
logger.info(
"Corrupted model file %s for %s. "
log_info_on_rank0(
logger,
f"Corrupted model file {base_name} for {model_name_or_path}. "
"Will selectively clean and re-download this file.",
base_name,
model_name_or_path,
)
# Selective cleanup for single file
_cleanup_corrupted_files_selective(model_name_or_path, [f])
return None
if len(local_weight_files) > 0:
logger.info(
"Found local HF snapshot for %s at %s; skipping download.",
model_name_or_path,
found_local_snapshot_dir,
log_info_on_rank0(
logger,
f"Found local HF snapshot for {model_name_or_path} at "
f"{found_local_snapshot_dir}; skipping download.",
)
return found_local_snapshot_dir
else:
logger.info(
"Local HF snapshot at %s has no files matching %s; will attempt download.",
found_local_snapshot_dir,
allow_patterns,
log_info_on_rank0(
logger,
f"Local HF snapshot at {found_local_snapshot_dir} has no files matching "
f"{allow_patterns}; will attempt download.",
)
return None