alihan/exo
1
0
Fork 0
mirror of https://github.com/exo-explore/exo.git synced 2025-10-23 02:57:14 +03:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #660 from exo-explore/robustdownload

Browse Source 
cleanup tmp files on failed download
This commit is contained in:
Alex Cheema
2025-01-30 20:25:15 +00:00
committed by GitHub
parent 90690a7d10 0bebf8dfde
commit 7a75fb09b2
1 changed files with 35 additions and 28 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
exo/download/new_shard_download.py
View File

@@ -75,7 +75,7 @@ async def fetch_file_list(repo_id, revision, path=""):
url = f"{api_url}/{path}" if path else api_url
headers = await get_auth_headers()
async with aiohttp.ClientSession(timeout=aiohttp.ClientTimeout(total=30, connect=10, sock_read=1800, sock_connect=60)) as session:
async with aiohttp.ClientSession(timeout=aiohttp.ClientTimeout(total=30, connect=10, sock_read=30, sock_connect=10)) as session:
async with session.get(url, headers=headers) as response:
if response.status == 200:
data = await response.json()
@@ -84,7 +84,7 @@ async def fetch_file_list(repo_id, revision, path=""):
if item["type"] == "file":
files.append({"path": item["path"], "size": item["size"]})
elif item["type"] == "directory":
subfiles = await fetch_file_list(session, repo_id, revision, item["path"])
subfiles = await fetch_file_list(repo_id, revision, item["path"])
files.extend(subfiles)
return files
else:
@@ -92,6 +92,8 @@ async def fetch_file_list(repo_id, revision, path=""):
@retry(stop=stop_after_attempt(5), wait=wait_exponential(multiplier=0.5))
async def download_file(repo_id: str, revision: str, path: str, target_dir: Path, on_progress: Callable[[int, int], None] = lambda _, __: None) -> Path:
temp_file_name = None
try:
if (target_dir/path).exists(): return target_dir/path
await aios.makedirs((target_dir/path).parent, exist_ok=True)
base_url = f"{get_hf_endpoint()}/{repo_id}/resolve/{revision}/"
@@ -103,9 +105,15 @@ async def download_file(repo_id: str, revision: str, path: str, target_dir: Path
length = int(r.headers.get('content-length', 0))
n_read = 0
async with aiofiles.tempfile.NamedTemporaryFile(dir=target_dir, delete=False) as temp_file:
temp_file_name = temp_file.name
while chunk := await r.content.read(1024 * 1024): on_progress(n_read := n_read + await temp_file.write(chunk), length)
await aios.rename(temp_file.name, target_dir/path)
return target_dir/path
finally:
if temp_file_name: # attempt to delete tmp file if it still exists
try: await aios.unlink(temp_file_name)
except: pass
def calculate_repo_progress(shard: Shard, repo_id: str, revision: str, file_progress: Dict[str, RepoFileProgressEvent], all_start_time: float) -> RepoProgressEvent:
all_total_bytes = sum([p.total for p in file_progress.values()])
@@ -233,4 +241,3 @@ class NewShardDownloader(ShardDownloader):
if DEBUG >= 6: print("Downloaded shards:", downloads)
if any(isinstance(d, Exception) for d in downloads) and DEBUG >= 1: print("Error downloading shards:", [d for d in downloads if isinstance(d, Exception)])
return [d for d in downloads if not isinstance(d, Exception)]