Update inference.py

inference.py

@@ -50,6 +50,7 @@ def get_args():
     parser.add_argument("--video_path", type=str, required=True, help="Path to a source video")
     parser.add_argument("--output_path", type=str, default="./outputs", help="Directory of output")
     parser.add_argument("--condition", type=str, default="depth", help="Condition of structure sequence")
+    parser.add_argument("--inference_steps", type=int, default=25, help="Number of inference steps")
     parser.add_argument("--video_length", type=int, default=15, help="Length of synthesized video")
     parser.add_argument("--height", type=int, default=512, help="Height of synthesized video, and should be a multiple of 32")
     parser.add_argument("--width", type=int, default=512, help="Width of synthesized video, and should be a multiple of 32")
@@ -122,13 +123,13 @@ if __name__ == "__main__":
     if args.is_long_video:
         window_size = int(np.sqrt(args.video_length))
         sample = pipe.generate_long_video(args.prompt + POS_PROMPT, video_length=args.video_length, frames=pil_annotation, 
-                    num_inference_steps=25, smooth_steps=args.smoother_steps, window_size=window_size,
+                    num_inference_steps=args.inference_steps, smooth_steps=args.smoother_steps, window_size=window_size,
                     generator=generator, guidance_scale=12.5, negative_prompt=NEG_PROMPT,
                     width=args.width, height=args.height
                 ).videos
     else:
         sample = pipe(args.prompt + POS_PROMPT, video_length=args.video_length, frames=pil_annotation, 
-                    num_inference_steps=50, smooth_steps=args.smoother_steps,
+                    num_inference_steps=args.inference_steps, smooth_steps=args.smoother_steps,
                     generator=generator, guidance_scale=12.5, negative_prompt=NEG_PROMPT,
                     width=args.width, height=args.height
                 ).videos