update

finetune/train_cogvideox_image_to_video_lora.py

@@ -1,5 +1,4 @@
-# Copyright 2024 The HuggingFace Team.
-# All rights reserved.
+# Copyright 2024 The CogView team, Tsinghua University & ZhipuAI and The HuggingFace Team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -45,10 +44,7 @@ from diffusers import (
 from diffusers.models.embeddings import get_3d_rotary_pos_embed
 from diffusers.optimization import get_scheduler
 from diffusers.pipelines.cogvideo.pipeline_cogvideox import get_resize_crop_region_for_grid
-from diffusers.training_utils import (
-    cast_training_params,
-    clear_objs_and_retain_memory,
-)
+from diffusers.training_utils import cast_training_params, free_memory
 from diffusers.utils import (
     check_min_version,
     convert_unet_state_dict_to_peft,
@@ -58,6 +54,10 @@ from diffusers.utils import (
 )
 from diffusers.utils.hub_utils import load_or_create_model_card, populate_model_card
 from diffusers.utils.torch_utils import is_compiled_module
+from torchvision.transforms.functional import center_crop, resize
+from torchvision.transforms import InterpolationMode
+import torchvision.transforms as TT
+import numpy as np
 
 
 if is_wandb_available():
@@ -236,6 +236,12 @@ def get_args():
         default=720,
         help="All input videos are resized to this width.",
     )
+    parser.add_argument(
+        "--video_reshape_mode",
+        type=str,
+        default="center",
+        help="All input videos are reshaped to this mode. Choose between ['center', 'random', 'none']",
+    )
     parser.add_argument("--fps", type=int, default=8, help="All input videos will be used at this FPS.")
     parser.add_argument(
         "--max_num_frames", type=int, default=49, help="All input videos will be truncated to these many frames."
@@ -442,6 +448,7 @@ class VideoDataset(Dataset):
         video_column: str = "video",
         height: int = 480,
         width: int = 720,
+        video_reshape_mode: str = "center",
         fps: int = 8,
         max_num_frames: int = 49,
         skip_frames_start: int = 0,
@@ -450,6 +457,7 @@ class VideoDataset(Dataset):
         id_token: Optional[str] = None,
     ) -> None:
         super().__init__()
+
         self.instance_data_root = Path(instance_data_root) if instance_data_root is not None else None
         self.dataset_name = dataset_name
         self.dataset_config_name = dataset_config_name
@@ -457,12 +465,14 @@ class VideoDataset(Dataset):
         self.video_column = video_column
         self.height = height
         self.width = width
+        self.video_reshape_mode = video_reshape_mode
         self.fps = fps
         self.max_num_frames = max_num_frames
         self.skip_frames_start = skip_frames_start
         self.skip_frames_end = skip_frames_end
         self.cache_dir = cache_dir
         self.id_token = id_token or ""
+
         if dataset_name is not None:
             self.instance_prompts, self.instance_video_paths = self._load_dataset_from_hub()
         else:
@@ -561,6 +571,38 @@ class VideoDataset(Dataset):
 
         return instance_prompts, instance_videos
 
+    def _resize_for_rectangle_crop(self, arr):
+            image_size = self.height, self.width
+            reshape_mode = self.video_reshape_mode
+            if arr.shape[3] / arr.shape[2] > image_size[1] / image_size[0]:
+                arr = resize(
+                    arr,
+                    size=[image_size[0], int(arr.shape[3] * image_size[0] / arr.shape[2])],
+                    interpolation=InterpolationMode.BICUBIC,
+                )
+            else:
+                arr = resize(
+                    arr,
+                    size=[int(arr.shape[2] * image_size[1] / arr.shape[3]), image_size[1]],
+                    interpolation=InterpolationMode.BICUBIC,
+                )
+
+            h, w = arr.shape[2], arr.shape[3]
+            arr = arr.squeeze(0)
+
+            delta_h = h - image_size[0]
+            delta_w = w - image_size[1]
+
+            if reshape_mode == "random" or reshape_mode == "none":
+                top = np.random.randint(0, delta_h + 1)
+                left = np.random.randint(0, delta_w + 1)
+            elif reshape_mode == "center":
+                top, left = delta_h // 2, delta_w // 2
+            else:
+                raise NotImplementedError
+            arr = TT.functional.crop(arr, top=top, left=left, height=image_size[0], width=image_size[1])
+            return arr
+
     def _preprocess_data(self):
         try:
             import decord
@@ -571,14 +613,15 @@ class VideoDataset(Dataset):
 
         decord.bridge.set_bridge("torch")
 
-        videos = []
-        train_transforms = transforms.Compose(
-            [
-                transforms.Lambda(lambda x: x / 255.0 * 2.0 - 1.0),
-            ]
+        progress_dataset_bar = tqdm(
+            range(0, len(self.instance_video_paths)),
+            desc="Loading progress resize and crop videos",
         )
 
+        videos = []
+
         for filename in self.instance_video_paths:
+            progress_dataset_bar.update(1)
             video_reader = decord.VideoReader(uri=filename.as_posix(), width=self.width, height=self.height)
             video_num_frames = len(video_reader)
 
@@ -605,9 +648,12 @@ class VideoDataset(Dataset):
             assert (selected_num_frames - 1) % 4 == 0
 
             # Training transforms
-            frames = frames.float()
-            frames = torch.stack([train_transforms(frame) for frame in frames], dim=0)
-            videos.append(frames.permute(0, 3, 1, 2).contiguous())  # [F, C, H, W]
+            frames = (frames - 127.5) / 127.5
+            frames = frames.permute(0, 3, 1, 2) # [F, C, H, W]
+            frames = self._resize_for_rectangle_crop(frames)
+            videos.append(frames.contiguous())  # [F, C, H, W]
+
+        progress_dataset_bar.close()
 
         return videos
 
@@ -727,7 +773,7 @@ def log_validation(
 
     videos = []
     for _ in range(args.num_validation_videos):
-        video = pipe(**pipeline_args, generator=generator, output_type="np").frames[0]
+        video = pipe(**pipeline_args, generator=generator, output_type="pil").frames[0]
         videos.append(video)
 
     for tracker in accelerator.trackers:
@@ -756,7 +802,8 @@ def log_validation(
                 }
             )
 
-    clear_objs_and_retain_memory([pipe])
+    del pipe
+    free_memory()
 
     return videos
 
@@ -1204,6 +1251,7 @@ def main(args):
         video_column=args.video_column,
         height=args.height,
         width=args.width,
+        video_reshape_mode=args.video_reshape_mode,
         fps=args.fps,
         max_num_frames=args.max_num_frames,
         skip_frames_start=args.skip_frames_start,
@@ -1212,7 +1260,8 @@ def main(args):
         id_token=args.id_token,
     )
 
-    def encode_video(video):
+    def encode_video(video, bar):
+        bar.update(1)
         video = video.to(accelerator.device, dtype=vae.dtype).unsqueeze(0)
         video = video.permute(0, 2, 1, 3, 4)  # [B, C, F, H, W]
         image = video[:, :, :1].clone()
@@ -1238,7 +1287,13 @@ def main(args):
         )
         for prompt in train_dataset.instance_prompts
     ]
-    train_dataset.instance_videos = [encode_video(video) for video in train_dataset.instance_videos]
+
+    progress_encode_bar = tqdm(
+        range(0, len(train_dataset.instance_videos)),
+        desc="Loading Encode videos",
+    )
+    train_dataset.instance_videos = [encode_video(video, progress_encode_bar) for video in train_dataset.instance_videos]
+    progress_encode_bar.close()
 
     def collate_fn(examples):
         videos = []
@@ -1378,9 +1433,6 @@ def main(args):
     )
     vae_scale_factor_spatial = 2 ** (len(vae.config.block_out_channels) - 1)
 
-    # Delete VAE and Text Encoder to save memory
-    clear_objs_and_retain_memory([vae, text_encoder])
-
     # For DeepSpeed training
     model_config = transformer.module.config if hasattr(transformer, "module") else transformer.config
 
@@ -1550,7 +1602,8 @@ def main(args):
         )
 
         # Cleanup trained models to save memory
-        clear_objs_and_retain_memory([transformer])
+        del transformer
+        free_memory()
 
         # Final test inference
         pipe = CogVideoXImageToVideoPipeline.from_pretrained(

finetune/train_cogvideox_lora.py

@@ -1,5 +1,4 @@
-# Copyright 2024 The HuggingFace Team.
-# All rights reserved.
+# Copyright 2024 The CogView team, Tsinghua University & ZhipuAI and The HuggingFace Team. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.