Merge branch 'main' of https://huggingface.co/spaces/PAIR/Text2Video-Zero into main

model.py

@@ -1,7 +1,7 @@
 from enum import Enum
 import gc
 import numpy as np
-import tomesd
+#import tomesd
 import torch
 
 from diffusers import StableDiffusionInstructPix2PixPipeline, StableDiffusionControlNetPipeline, ControlNetModel, UNet2DConditionModel, TextToVideoZeroPipeline
@@ -83,15 +83,16 @@ class Model:
                          generator=self.generator,
                          **kwargs)
 
-    def inference(self, split_to_chunks=False, chunk_size=8, **kwargs):
+    def inference(self, split_to_chunks=False, chunk_size=2, **kwargs):
         if not hasattr(self, "pipe") or self.pipe is None:
             return
-
+        '''
         if "merging_ratio" in kwargs:
             merging_ratio = kwargs.pop("merging_ratio")
 
             # if merging_ratio > 0:
             tomesd.apply_patch(self.pipe, ratio=merging_ratio)
+        '''
         seed = kwargs.pop('seed', 0)
         if seed < 0:
             seed = self.generator.seed()
@@ -134,7 +135,7 @@ class Model:
     def process_controlnet_canny(self,
                                  video_path,
                                  prompt,
-                                 chunk_size=8,
+                                 chunk_size=2,
                                  watermark='Picsart AI Research',
                                  merging_ratio=0.0,
                                  num_inference_steps=20,
@@ -200,7 +201,7 @@ class Model:
     def process_controlnet_depth(self,
                                  video_path,
                                  prompt,
-                                 chunk_size=8,
+                                 chunk_size=2,
                                  watermark='Picsart AI Research',
                                  merging_ratio=0.0,
                                  num_inference_steps=20,
@@ -263,7 +264,7 @@ class Model:
     def process_controlnet_pose(self,
                                 video_path,
                                 prompt,
-                                chunk_size=8,
+                                chunk_size=2,
                                 watermark='Picsart AI Research',
                                 merging_ratio=0.0,
                                 num_inference_steps=20,
@@ -326,7 +327,7 @@ class Model:
                                     db_path,
                                     video_path,
                                     prompt,
-                                    chunk_size=8,
+                                    chunk_size=2,
                                     watermark='Picsart AI Research',
                                     merging_ratio=0.0,
                                     num_inference_steps=20,
@@ -397,7 +398,7 @@ class Model:
                         start_t=0,
                         end_t=-1,
                         out_fps=-1,
-                        chunk_size=8,
+                        chunk_size=2,
                         watermark='Picsart AI Research',
                         merging_ratio=0.0,
                         use_cf_attn=True,
@@ -434,7 +435,7 @@ class Model:
                            t0=44,
                            t1=47,
                            n_prompt="",
-                           chunk_size=8,
+                           chunk_size=2,
                            video_length=8,
                            watermark='Picsart AI Research',
                            merging_ratio=0.0,