Update app.py

app.py

@@ -91,12 +91,13 @@ def generate(segment, image, prompt, size, guidance_scale, num_inference_steps,
         #)
         pipe.text_encoder.to("cuda")
         pipe.text_encoder_2.to("cuda")
-        #with torch.no_grad():
-        prompt_embeds, prompt_attention_mask, negative_prompt_embeds, negative_attention_mask, pooled_prompt_embeds, negative_pooled_prompt_embeds = pipe.encode_prompt(
-            prompt=prompt, do_classifier_free_guidance=True, negative_prompt=negative_prompt, device=device
-        )
+        with torch.no_grad():
+            prompt_embeds, prompt_attention_mask, negative_prompt_embeds, negative_attention_mask, pooled_prompt_embeds, negative_pooled_prompt_embeds = pipe.encode_prompt(
+                prompt=prompt, do_classifier_free_guidance=True, negative_prompt=negative_prompt, device=device
+            )
         pipe.text_encoder.to("cpu")
         pipe.text_encoder_2.to("cpu")
+        torch.cuda.empty_cache()
         generator = torch.Generator(device='cuda').manual_seed(seed)
         transformer_dtype = pipe.transformer.dtype
         prompt_embeds = prompt_embeds.to(transformer_dtype)
@@ -133,6 +134,7 @@ def generate(segment, image, prompt, size, guidance_scale, num_inference_steps,
         )
         image_latents = image_latents.to(pipe.transformer.dtype)
         pipe.vae.to("cpu")
+        torch.cuda.empty_cache()
         guidance = torch.tensor([guidance_scale] * latents.shape[0], dtype=transformer_dtype, device=device) * 1000.0
     else:
         state_file = f"rv_L_{segment-1}_{seed}.pt"