zynapse commited on
Commit
98aa6bd
·
verified ·
1 Parent(s): 67bd57c

Upload app.py

Browse files
Files changed (1) hide show
  1. app.py +64 -6
app.py CHANGED
@@ -12,6 +12,7 @@ import torchvision.transforms as transforms
12
  from PIL import Image
13
  import gradio as gr
14
  import timm
 
15
  try:
16
  import joblib
17
  except ImportError:
@@ -479,12 +480,7 @@ class PlantClassifier:
479
  url = config['drive_weights_url']
480
  local_path = os.path.join(self.models_dir, 'dinov3_backbone_from_drive.pth')
481
  print(f"⬇️ Downloading backbone weights from {url} -> {local_path}")
482
- r = requests.get(url, stream=True, timeout=60)
483
- r.raise_for_status()
484
- with open(local_path, 'wb') as f:
485
- for chunk in r.iter_content(chunk_size=8192):
486
- if chunk:
487
- f.write(chunk)
488
  base_obj = torch.load(local_path, map_location=self.device, weights_only=False)
489
  if isinstance(base_obj, nn.Module):
490
  backbone = base_obj
@@ -635,6 +631,68 @@ class PlantClassifier:
635
  'actual_num_classes': num_classes
636
  }
637
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
638
  def get_transform(self, input_size, preproc=None):
639
  """Get image preprocessing transforms"""
640
  if preproc == 'dann224':
 
12
  from PIL import Image
13
  import gradio as gr
14
  import timm
15
+ import re
16
  try:
17
  import joblib
18
  except ImportError:
 
480
  url = config['drive_weights_url']
481
  local_path = os.path.join(self.models_dir, 'dinov3_backbone_from_drive.pth')
482
  print(f"⬇️ Downloading backbone weights from {url} -> {local_path}")
483
+ self._download_google_drive_file(url, local_path)
 
 
 
 
 
484
  base_obj = torch.load(local_path, map_location=self.device, weights_only=False)
485
  if isinstance(base_obj, nn.Module):
486
  backbone = base_obj
 
631
  'actual_num_classes': num_classes
632
  }
633
 
634
+ def _extract_drive_file_id(self, url: str):
635
+ """Extract Google Drive file id from various URL forms."""
636
+ # Try query param id=...
637
+ m = re.search(r"[?&]id=([^&]+)", url)
638
+ if m:
639
+ return m.group(1)
640
+ # Try /d/<id>/ pattern
641
+ m = re.search(r"/d/([A-Za-z0-9_-]+)", url)
642
+ if m:
643
+ return m.group(1)
644
+ return None
645
+
646
+ def _download_google_drive_file(self, url: str, destination: str):
647
+ """Download a file from Google Drive handling confirmation for large files.
648
+
649
+ Falls back to raising if the final response doesn't look like a binary download.
650
+ """
651
+ if requests is None:
652
+ raise RuntimeError("requests not installed")
653
+ session = requests.Session()
654
+ file_id = self._extract_drive_file_id(url)
655
+ if not file_id:
656
+ # Assume direct binary link
657
+ with session.get(url, stream=True, timeout=60) as r:
658
+ r.raise_for_status()
659
+ self._stream_to_file(r, destination)
660
+ return
661
+ # Step 1: initial request
662
+ base = "https://drive.google.com/uc?export=download"
663
+ params = {"id": file_id, "acknowledgeAbuse": "true"}
664
+ with session.get(base, params=params, stream=True, timeout=60) as r1:
665
+ if "content-disposition" in r1.headers:
666
+ self._stream_to_file(r1, destination)
667
+ return
668
+ # look for confirm token
669
+ text = r1.text or ""
670
+ m = re.search(r"confirm=([0-9A-Za-z_]+)", text)
671
+ token = m.group(1) if m else None
672
+ if not token:
673
+ # Sometimes token lives in cookies
674
+ for k, v in r1.cookies.items():
675
+ if k.startswith("download_warning"):
676
+ token = v
677
+ break
678
+ if not token:
679
+ # No token and no disposition -> not a file
680
+ raise RuntimeError("Google Drive download requires confirmation but token was not found.")
681
+ # Step 2: confirmed request
682
+ params["confirm"] = token
683
+ with session.get(base, params=params, stream=True, timeout=60) as r2:
684
+ r2.raise_for_status()
685
+ if "content-disposition" not in r2.headers and "octet-stream" not in r2.headers.get("content-type", ""):
686
+ raise RuntimeError("Google Drive returned non-binary response after confirmation.")
687
+ self._stream_to_file(r2, destination)
688
+
689
+ def _stream_to_file(self, response, destination: str):
690
+ os.makedirs(os.path.dirname(destination), exist_ok=True)
691
+ with open(destination, 'wb') as f:
692
+ for chunk in response.iter_content(chunk_size=1048576):
693
+ if chunk:
694
+ f.write(chunk)
695
+
696
  def get_transform(self, input_size, preproc=None):
697
  """Get image preprocessing transforms"""
698
  if preproc == 'dann224':