Upload folder using huggingface_hub

.github/workflows/deploy-to-hf-spaces.yml

@@ -0,0 +1,57 @@
+name: Deploy to HuggingFace Spaces
+
+on:
+  push:
+    branches:
+      - main
+
+jobs:
+  deploy:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+          lfs: true
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.10'
+
+      - name: Install huggingface_hub
+        run: pip install huggingface_hub
+
+      - name: Deploy to HuggingFace Spaces
+        env:
+          HF_TOKEN: ${{ secrets.HF_TOKEN }}
+        run: |
+          python - <<EOF
+          from huggingface_hub import HfApi
+          import os
+
+          api = HfApi(token=os.environ['HF_TOKEN'])
+
+          # Upload files to the Space
+          api.upload_folder(
+              folder_path='.',
+              repo_id='eduardofarina/MedGemma1.5ReportGenerator',
+              repo_type='space',
+              ignore_patterns=[
+                  'models/*',
+                  '.git/*',
+                  '.github/*',
+                  '__pycache__/*',
+                  '*.pyc',
+                  '.env',
+                  'environment.yml',
+                  'setup*.bat',
+                  'setup*.sh',
+                  'check_gpu.py',
+                  'download_model.py',
+                  'dicom_utils.py',
+              ],
+          )
+          print('Deployment to HuggingFace Spaces complete!')
+          EOF

model_handler.py

@@ -86,10 +86,10 @@ class MedGemmaHandler:
 
         # Load model with proper device configuration
         if self.device == "cuda" and torch.cuda.is_available():
-            print("Loading model on GPU with bfloat16...")
+            print("Loading model on GPU with float16...")
             self.model = AutoModelForImageTextToText.from_pretrained(
                 self.model_id,
-                torch_dtype=torch.bfloat16,
+                torch_dtype=torch.float16,
                 device_map="cuda",
                 token=hf_token,
             )
@@ -139,7 +139,7 @@ class MedGemmaHandler:
 
         # Move to device with proper dtype
         if self.device == "cuda":
-            inputs = inputs.to(self.model.device, dtype=torch.bfloat16)
+            inputs = inputs.to(self.model.device, dtype=torch.float16)
         else:
             inputs = inputs.to(self.model.device)