Update pipeline.py

pipeline.py

@@ -1006,36 +1006,25 @@ class AnimateDiffPipeline(DiffusionPipeline, TextualInversionLoaderMixin, IPAdap
                     # set the step index to the current batch
                     self.scheduler._step_index = i
                     
-                    print(f"current_context_latents shape: {current_context_latents.shape}")
-                    print(f"noise_pred shape: {noise_pred.shape}")
-                        
-                        
-                    # print size of sigmas
-                    print(f"sigma len: {len(self.scheduler.sigmas)}")
-                    
-                    # print step index
-                    print(f"step index: {self.scheduler.step_index}")
-                    print(f"sigma: {self.scheduler.sigmas[self.scheduler.step_index + 1]}")
-                    
                     # compute the previous noisy sample x_t -> x_t-1
                     current_context_latents = self.scheduler.step(noise_pred, t, current_context_latents, **extra_step_kwargs).prev_sample
                     
                     # remove the appended frames from the end of the current_context_latents
-                    if wrap_count > 0:
-                        # remove the ending frames from current_context_latents
-                        current_context_latents = current_context_latents[:, :, :-wrap_count, :, :]
-                        # remove the ending frames from noise_pred
-                        noise_pred = noise_pred[:, :, :-wrap_count, :, :]
-                        # print the shape of the current_context_latents and noise_pred
-                    
-                    # # if context_start + context_size > num_frames: remove the appended frames from the end of the current_context_latents
                     # if wrap_count > 0:
-                    #     # add the ending frames from current_context_latents to the start of the latent_sum
-                    #     latent_sum[:, :, 0:wrap_count, :, :] += current_context_latents[:, :, -wrap_count:, :, :]
-                    #     # increase the counter for the ending frames
-                    #     latent_counter[0:wrap_count] += 1
                     #     # remove the ending frames from current_context_latents
                     #     current_context_latents = current_context_latents[:, :, :-wrap_count, :, :]
+                    #     # remove the ending frames from noise_pred
+                    #     noise_pred = noise_pred[:, :, :-wrap_count, :, :]
+                    #     # print the shape of the current_context_latents and noise_pred
+                    
+                    # if context_start + context_size > num_frames: remove the appended frames from the end of the current_context_latents
+                    if wrap_count > 0:
+                        # add the ending frames from current_context_latents to the start of the latent_sum
+                        latent_sum[:, :, 0:wrap_count, :, :] += current_context_latents[:, :, -wrap_count:, :, :]
+                        # increase the counter for the ending frames
+                        latent_counter[0:wrap_count] += 1
+                        # remove the ending frames from current_context_latents
+                        current_context_latents = current_context_latents[:, :, :-wrap_count, :, :]
 
                     #add the context current_context_latents back to the latent sum starting from the current context start
                     latent_sum[:, :, current_context_start : current_context_start + context_size, :, :] += current_context_latents