Add HuggingFace token support for Sesame CSM

- Update GitHub Actions to use HUGGINGFACE_TOKEN secret
- Gracefully handle missing token with helpful error messages
- Add proper token authentication to model loading
- Provide clear setup instructions for model access

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

Author: YedPool

.github/workflows/test-docker.yml

@@ -365,12 +365,21 @@ jobs:
           exit 1
         fi
         
-    - name: Test Sesame TTS container startup (without model loading)
+    - name: Test Sesame TTS container startup
       run: |
-        # Start Sesame TTS container (will fail without HF token but should start server)
-        docker run -d --name sesame-tts-test \
-          -p 8001:8001 \
-          sesame-tts:test
+        # Start Sesame TTS container with HuggingFace token if available
+        if [ -n "${{ secrets.HUGGINGFACE_TOKEN }}" ]; then
+          echo "🔑 Using HuggingFace token for model access"
+          docker run -d --name sesame-tts-test \
+            -p 8001:8001 \
+            -e HUGGINGFACE_HUB_TOKEN=${{ secrets.HUGGINGFACE_TOKEN }} \
+            sesame-tts:test
+        else
+          echo "⚠️ No HuggingFace token available - testing container startup only"
+          docker run -d --name sesame-tts-test \
+            -p 8001:8001 \
+            sesame-tts:test
+        fi
           
         # Show immediate logs
         echo "📋 Initial Sesame TTS logs:"

sesame-tts/main.py

@@ -75,19 +75,34 @@ def load_model(self):
             model_id = "sesame/csm-1b"
             logger.info(f"Loading Sesame CSM model: {model_id}")
 
+            # Check for HuggingFace token
+            hf_token = os.getenv("HUGGINGFACE_HUB_TOKEN")
+            if not hf_token:
+                logger.warning("No HUGGINGFACE_HUB_TOKEN environment variable found")
+                logger.warning("You may need to accept the model license and set your HF token")
+            else:
+                logger.info("✓ HuggingFace token found")
+            
             # Load processor and model
-            self.processor = AutoProcessor.from_pretrained(model_id)
+            self.processor = AutoProcessor.from_pretrained(
+                model_id,
+                token=hf_token
+            )
             self.model = CsmForConditionalGeneration.from_pretrained(
                 model_id,
                 device_map=self.device,
-                torch_dtype=torch.float16 if self.device == "cuda" else torch.float32
+                torch_dtype=torch.float16 if self.device == "cuda" else torch.float32,
+                token=hf_token
             )
 
             logger.info("✓ Sesame CSM model loaded successfully")
             return True
 
         except Exception as e:
             logger.error(f"Failed to load model: {e}")
+            logger.error("Make sure you have:")
+            logger.error("1. Accepted the license at https://huggingface.co/sesame/csm-1b")
+            logger.error("2. Set HUGGINGFACE_HUB_TOKEN environment variable")
             return False
 
     def generate_audio(self, text: str, speaker: int = 0, max_audio_length_ms: int = 10000):