fix: free GPU memory between samples to prevent VRAM fragmentation

All samples within a single generate call share one @spaces.GPU
reservation. Without explicit cleanup, each sample's intermediate
tensors accumulate in the CUDA allocator cache, fragmenting VRAM and
causing progressive quality degradation on samples 2, 3, 4+.

torch.cuda.empty_cache() after each sample flushes the allocator so
every sample starts from a clean memory state, making quality
consistent across all generations.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

app.py

@@ -535,6 +535,11 @@ def _taro_gpu_infer(video_file, seed_val, cfg_scale, num_steps, mode,
                     _TARO_INFERENCE_CACHE.pop(next(iter(_TARO_INFERENCE_CACHE)))
             results.append((wavs, cavp_feats, onset_feats))
 
+        # Free GPU memory between samples so VRAM fragmentation doesn't
+        # degrade diffusion quality on samples 2, 3, 4, etc.
+        if torch.cuda.is_available():
+            torch.cuda.empty_cache()
+
     return results
 
 # Attach a context slot for the CPU wrapper to pass pre-computed data