Upload pipeline_ltx_video.py

LTX-Video/ltx_video/pipelines/pipeline_ltx_video.py

@@ -1206,6 +1206,11 @@ class LTXVideoPipeline(DiffusionPipeline):
                         orig_conditioning_mask,
                         generator,
                     )
+                    
+                try:
+                   print(f"[LTX6]LATENTS {latents.shape}")
+                except Exception:
+                   pass
 
                 latent_model_input = (
                     torch.cat([latents] * num_conds) if num_conds > 1 else latents
@@ -1213,6 +1218,11 @@ class LTXVideoPipeline(DiffusionPipeline):
                 latent_model_input = self.scheduler.scale_model_input(
                     latent_model_input, t
                 )
+                
+                try:
+                   print(f"[LTX7]LATENTS {latent_model_input.shape}")
+                except Exception:
+                   pass
 
                 current_timestep = t
                 if not torch.is_tensor(current_timestep):
@@ -1328,6 +1338,11 @@ class LTXVideoPipeline(DiffusionPipeline):
                     extra_step_kwargs,
                     stochastic_sampling=stochastic_sampling,
                 )
+                
+                try:
+                   print(f"[LTX8]LATENTS {latents.shape}")
+                except Exception:
+                   pass
 
                 # call the callback, if provided
                 if i == len(timesteps) - 1 or (
@@ -1341,7 +1356,7 @@ class LTXVideoPipeline(DiffusionPipeline):
 
 
         try:
-            print(f"[LTX6]LATENTS {latents.shape}")
+            print(f"[LTX9]LATENTS {latents.shape}")
         except Exception:
             pass
             
@@ -1352,6 +1367,12 @@ class LTXVideoPipeline(DiffusionPipeline):
 
         # Remove the added conditioning latents
         latents = latents[:, num_cond_latents:]
+        
+        
+        try:
+             print(f"[LTX10]LATENTS {latents.shape}")
+        except Exception:
+            pass
 
         latents = self.patchifier.unpatchify(
             latents=latents,
@@ -1379,7 +1400,8 @@ class LTXVideoPipeline(DiffusionPipeline):
                 )
             else:
                 decode_timestep = None
-            latents = self.tone_map_latents(latents, tone_map_compression_ratio)
+            latents = self.tone_map_latents(latents, tone_map
+            _compression_ratio)
             image = vae_decode(
                 latents,
                 self.vae,
@@ -1387,6 +1409,11 @@ class LTXVideoPipeline(DiffusionPipeline):
                 vae_per_channel_normalize=kwargs["vae_per_channel_normalize"],
                 timestep=decode_timestep,
             )
+            
+            try:
+                 print(f"[LTX11]LATENTS {latents.shape}")
+            except Exception:
+                 pass
 
             image = self.image_processor.postprocess(image, output_type=output_type)