fix: address review feedback on CUDA backend swap

- Use YAML block scalar for inline run with colons (build-cuda.yml)
- Explicitly set VOICEBOX_BACKEND_VARIANT=cpu instead of setdefault (server.py)
- Use Path.replace() for atomic move on all platforms (cuda_download.py)
- Log actual exception in checksum fetch warning (cuda_download.py)

.github/workflows/build-cuda.yml

@@ -33,7 +33,8 @@ jobs:
           pip install torchaudio --index-url https://download.pytorch.org/whl/cu121
 
       - name: Verify CUDA support in torch
-        run: python -c "import torch; print(f'CUDA available in build: {torch.cuda.is_available()}'); print(f'CUDA version: {torch.version.cuda}')"
+        run: |
+          python -c "import torch; print(f'CUDA available in build: {torch.cuda.is_available()}'); print(f'CUDA version: {torch.version.cuda}')"
 
       - name: Build CUDA server binary
         shell: bash

backend/cuda_download.py

@@ -126,8 +126,8 @@ async def download_cuda_binary(version: Optional[str] = None):
                     # Format: "sha256hex  filename\n"
                     expected_sha = sha_resp.text.strip().split()[0]
                     logger.info(f"Expected SHA-256: {expected_sha[:16]}...")
-            except Exception:
+            except Exception as e:
-                logger.warning("Could not fetch checksum file — skipping verification")
+                logger.warning(f"Could not fetch checksum file — skipping verification: {e}")
 
             # Download and concatenate parts
             total_downloaded = 0
@@ -169,10 +169,8 @@ async def download_cuda_binary(version: Optional[str] = None):
                 )
             logger.info(f"Integrity verified: {actual[:16]}...")
 
-        # Atomic move into place
+        # Atomic move into place (replace handles existing target on all platforms)
-        if final_path.exists():
+        temp_path.replace(final_path)
-            final_path.unlink()
-        temp_path.rename(final_path)
 
         # Make executable on Unix
         if sys.platform != "win32":

backend/server.py

@@ -84,8 +84,8 @@ if __name__ == "__main__":
             os.environ["VOICEBOX_BACKEND_VARIANT"] = "cuda"
             logger.info("Backend variant: CUDA")
         else:
-            os.environ.setdefault("VOICEBOX_BACKEND_VARIANT", "cpu")
+            os.environ["VOICEBOX_BACKEND_VARIANT"] = "cpu"
-            logger.info(f"Backend variant: {os.environ['VOICEBOX_BACKEND_VARIANT']}")
+            logger.info("Backend variant: CPU")
 
         logger.info(f"Parsed arguments: host={args.host}, port={args.port}, data_dir={args.data_dir}")