update: remove refiner

app.py

@@ -22,22 +22,22 @@ if gr.NO_RELOAD:
         variant="fp16", 
         use_safetensors=True,
     )
-    refiner = DiffusionPipeline.from_pretrained(
-        "stabilityai/stable-diffusion-xl-refiner-1.0",
-        text_encoder_2=base.text_encoder_2,
-        vae=base.vae,
-        torch_dtype=torch.float16,
-        use_safetensors=True,
-        variant="fp16",
-    )
+    # refiner = DiffusionPipeline.from_pretrained(
+    #     "stabilityai/stable-diffusion-xl-refiner-1.0",
+    #     text_encoder_2=base.text_encoder_2,
+    #     vae=base.vae,
+    #     torch_dtype=torch.float16,
+    #     use_safetensors=True,
+    #     variant="fp16",
+    # )
     pipeline = I2VGenXLPipeline.from_pretrained("ali-vilab/i2vgen-xl", torch_dtype=torch.float16, variant="fp16")
 
     base.to("cuda")
-    refiner.to("cuda")
+    # refiner.to("cuda")
     pipeline.to("cuda")
 
     base.unet = torch.compile(base.unet, mode="reduce-overhead", fullgraph=True)
-    refiner.unet = torch.compile(refiner.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)
 
 def generate(prompt: str, progress=gr.Progress()):
@@ -55,18 +55,18 @@ def generate(prompt: str, progress=gr.Progress()):
         ),
     ).images[0]
     progress((n_sdxl_steps * high_noise_frac, total_steps), desc="Refining first frame...")
-    image = refiner(
-        prompt=prompt,
-        num_inference_steps=n_sdxl_steps,
-        denoising_start=high_noise_frac,
-        image=image,
-        callback_on_step_end=create_progress_updater(
-            start=n_sdxl_steps * high_noise_frac,
-            total=total_steps,
-            desc="Refining first frame...",
-            progress=progress,
-        ),
-    ).images[0]
+    # image = refiner(
+    #     prompt=prompt,
+    #     num_inference_steps=n_sdxl_steps,
+    #     denoising_start=high_noise_frac,
+    #     image=image,
+    #     callback_on_step_end=create_progress_updater(
+    #         start=n_sdxl_steps * high_noise_frac,
+    #         total=total_steps,
+    #         desc="Refining first frame...",
+    #         progress=progress,
+    #     ),
+    # ).images[0]
     image = to_tensor(image)
     progress((n_sdxl_steps, total_steps), desc="Generating video...")
     frames: list[Image.Image] = pipeline(