auto resume

pull/529/head
lijiaxing 2023-12-07 10:19:48 +08:00
parent d0d39fa3ef
commit 1da080a58e
2 changed files with 9 additions and 3 deletions

View File

@ -902,6 +902,14 @@ now step_count is {train_state.step_count}",
latest_ckpt, step = self.query_latest_snapshot_step_boto3()
if latest_ckpt and not latest_ckpt.startswith("boto3:"):
latest_ckpt = ":".join(["boto3", latest_ckpt])
elif backend == "oss2":
latest_ckpt, step = self.query_latest_snapshot_step_boto3()
if latest_ckpt and not latest_ckpt.startswith("oss2:"):
latest_ckpt = ":".join(["oss2", latest_ckpt])
elif backend == "volc":
latest_ckpt, step = self.query_latest_snapshot_step_boto3()
if latest_ckpt and not latest_ckpt.startswith("volc:"):
latest_ckpt = ":".join(["volc", latest_ckpt])
elif backend == "local":
latest_ckpt, step = self.query_latest_snapshot_step_local()
if latest_ckpt and not latest_ckpt.startswith("local:"):

View File

@ -739,10 +739,8 @@ class AliClient(StorageClient):
if AliClient.is_fp_exists(handler, fp):
folder_name_list = []
for obj in handler.handler.ObjectIteratorV2(handler.client, prefix=fp):
folder_name_list.append(obj.key.split("/")[-1])
folder_name_list.append(obj.key.split(fp, maxsplit=1)[1].strip("/").split("/", maxsplit=1)[0])
return list(set(folder_name_list))
else:
if is_rank_for_log():
logger.warning(f"'{fp}' not found!")