Skip to content

Open-Sora: Democratizing Efficient Video Production for All

Python

fix dataset loading

Shen Chenhuib5a3334cafa9
diff --git a/configs/vae_magvit_v2/train/16x128x128_pixabay_8_GPU_pipeline.py b/configs/vae_magvit_v2/train/16x128x128_pixabay_8_GPU_pipeline.py
index 31af832..7b2c60c 100644
--- a/configs/vae_magvit_v2/train/16x128x128_pixabay_8_GPU_pipeline.py
+++ b/configs/vae_magvit_v2/train/16x128x128_pixabay_8_GPU_pipeline.py
@@ -4,6 +4,7 @@ dataset = dict(
num_frames=16,
frame_interval=3,
image_size=(128, 128),
+ get_text=False,
)
# Define acceleration
diff --git a/opensora/datasets/datasets.py b/opensora/datasets/datasets.py
index 4c2b316..d320ff7 100644
--- a/opensora/datasets/datasets.py
+++ b/opensora/datasets/datasets.py
@@ -29,6 +29,7 @@ class VideoTextDataset(torch.utils.data.Dataset):
frame_interval=1,
image_size=(256, 256),
transform_name="center",
+ get_text=True,
):
self.data_path = data_path
self.data = read_file(data_path)
@@ -39,6 +40,7 @@ class VideoTextDataset(torch.utils.data.Dataset):
"image": get_transforms_image(transform_name, image_size),
"video": get_transforms_video(transform_name, image_size),
}
+ self.get_text = get_text
def _print_data_number(self):
num_videos = 0
@@ -61,7 +63,7 @@ class VideoTextDataset(torch.utils.data.Dataset):
def getitem(self, index):
sample = self.data.iloc[index]
path = sample["path"]
- text = sample["text"]
+ text = sample["text"] if self.get_text else None
file_type = self.get_type(path)
if file_type == "video":
@@ -87,7 +89,11 @@ class VideoTextDataset(torch.utils.data.Dataset):
# TCHW -> CTHW
video = video.permute(1, 0, 2, 3)
- return {"video": video, "text": text}
+
+ if self.get_text:
+ return {"video": video, "text": text}
+ else:
+ return {"video": video}
def __getitem__(self, index):
for _ in range(10):