Earlier

app.py

@@ -175,6 +175,7 @@ def worker(input_image, prompt, n_prompt, seed, total_second_length, latent_wind
         image_encoder_last_hidden_state = image_encoder_output.last_hidden_state
 
         # Dtype
+        [llama_vec, clip_l_pooler, llama_vec_n, clip_l_pooler_n, llama_attention_mask, llama_attention_mask_n] = prompt_parameters.pop(0)
 
         image_encoder_last_hidden_state = image_encoder_last_hidden_state.to(transformer.dtype)
 
@@ -190,9 +191,6 @@ def worker(input_image, prompt, n_prompt, seed, total_second_length, latent_wind
         history_latents = torch.cat([history_latents, start_latent.to(history_latents)], dim=2)
         total_generated_latent_frames = 1
 
-        if len(prompt_parameters) > 1:
-            [llama_vec, clip_l_pooler, llama_vec_n, clip_l_pooler_n, llama_attention_mask, llama_attention_mask_n] = prompt_parameters.pop(0)
-
         for section_index in range(total_latent_sections):
             if stream.input_queue.top() == 'end':
                 stream.output_queue.push(('end', None))