fix: out of memory error

app.py

@@ -32,13 +32,14 @@ if gr.NO_RELOAD:
     # )
     pipeline = I2VGenXLPipeline.from_pretrained("ali-vilab/i2vgen-xl", torch_dtype=torch.float16, variant="fp16")
 
-    base.to("cuda")
+    # base.to("cuda")
     # refiner.to("cuda")
     # pipeline.to("cuda")
 
-    base.unet = torch.compile(base.unet, mode="reduce-overhead", fullgraph=True)
+    # base.unet = torch.compile(base.unet, mode="reduce-overhead", fullgraph=True)
     # refiner.unet = torch.compile(refiner.unet, mode="reduce-overhead", fullgraph=True)
     # pipeline.unet = torch.compile(pipeline.unet, mode="reduce-overhead", fullgraph=True)
+    base.enable_model_cpu_offload()
     pipeline.enable_model_cpu_offload()
     pipeline.unet.enable_forward_chunking()
 
@@ -77,12 +78,12 @@ def generate(prompt: str, progress=gr.Progress()):
         guidance_scale=9.0,
         generator=generator,
         decode_chunk_size=2,
-        num_frames=32,
+        num_frames=16,
     ).frames[0]
     progress((total_steps - 1, total_steps), desc="Finalizing...")
     frames = [to_tensor(frame.convert("RGB")).mul(255).byte().permute(1, 2, 0) for frame in frames]
     frames = torch.stack(frames)
-    torchvision.io.write_video("video.mp4", frames, fps=16)
+    torchvision.io.write_video("video.mp4", frames, fps=8)
     return "video.mp4"
 
 app = gr.Interface(