.

options/Video.py

@@ -1,4 +1,4 @@
 from .Video_model import Model
 
 def Video(image):
-    return Model.Video(image)
+    return Model.Video(image)[0]

options/Video_model/Model.py

@@ -53,8 +53,9 @@ def Video(
     base_count = len(glob(os.path.join(output_folder, "*.mp4")))
     video_path = os.path.join(output_folder, f"{base_count:06d}.mp4")
 
-    # Only use autocast if on CUDA, otherwise run without it on CPU
+    # Perform computation with appropriate dtype based on device
     if device == "cuda":
+        # Use float16 for GPU
         with torch.autocast(device_type='cuda', dtype=torch.float16):
             frames = pipeline(
                 image, height=height, width=width,
@@ -67,20 +68,21 @@ def Video(
                 generator=generator,
             ).frames[0]
     else:
-        # No autocast for CPU since it doesn't support float32 in autocast
-        frames = pipeline(
-            image, height=height, width=width,
-            num_inference_steps=num_inference_steps,
-            min_guidance_scale=min_guidance_scale,
-            max_guidance_scale=max_guidance_scale,
-            num_frames=num_frames, fps=fps, motion_bucket_id=motion_bucket_id,
-            decode_chunk_size=8,
-            noise_aug_strength=0.02,
-            generator=generator,
-        ).frames[0]
+        # Use float32 for CPU to avoid issues with half precision
+        with torch.autocast(device_type='cpu', dtype=torch.float32):
+            frames = pipeline(
+                image, height=height, width=width,
+                num_inference_steps=num_inference_steps,
+                min_guidance_scale=min_guidance_scale,
+                max_guidance_scale=max_guidance_scale,
+                num_frames=num_frames, fps=fps, motion_bucket_id=motion_bucket_id,
+                decode_chunk_size=8,
+                noise_aug_strength=0.02,
+                generator=generator,
+            ).frames[0]
 
     # Save the generated video
     save_video(frames, video_path, fps=fps, quality=5.0)
     torch.manual_seed(seed)
 
-    return video_path, seed
+    return video_path, seed