Ji4chenLi
commited on
Commit
•
979e017
1
Parent(s):
9ada050
update description
Browse files
app.py
CHANGED
|
@@ -20,15 +20,18 @@ from pipeline.t2v_turbo_vc2_pipeline import T2VTurboVC2Pipeline
|
|
| 20 |
DESCRIPTION = """# T2V-Turbo 🚀
|
| 21 |
|
| 22 |
Our model is distilled from [VideoCrafter2](https://ailab-cvc.github.io/videocrafter2/).
|
|
|
|
| 23 |
T2V-Turbo learns a LoRA on top of the base model by aligning to the reward feedback from [HPSv2.1](https://github.com/tgxs002/HPSv2/tree/master) and [InternVid2 Stage 2 Model](https://huggingface.co/OpenGVLab/InternVideo2-Stage2_1B-224p-f4).
|
|
|
|
| 24 |
T2V-Turbo-v2 optimizes the training techniques by finetuning the full base model and further aligns to [CLIPScore](https://huggingface.co/laion/CLIP-ViT-H-14-laion2B-s32B-b79K)
|
| 25 |
|
| 26 |
T2V-Turbo trains on pure WebVid-10M data, whereas T2V-Turbo-v2 carufully optimizes different learning objectives with a mixutre of VidGen-1M and WebVid-10M data.
|
| 27 |
|
| 28 |
Moreover, T2V-Turbo-v2 supports to distill motion priors from the training videos.
|
| 29 |
|
| 30 |
-
[Project page for T2V-Turbo](https://t2v-turbo.github.io)
|
| 31 |
-
|
|
|
|
| 32 |
"""
|
| 33 |
if torch.cuda.is_available():
|
| 34 |
DESCRIPTION += "\n<p>Running on CUDA 😀</p>"
|
|
|
|
| 20 |
DESCRIPTION = """# T2V-Turbo 🚀
|
| 21 |
|
| 22 |
Our model is distilled from [VideoCrafter2](https://ailab-cvc.github.io/videocrafter2/).
|
| 23 |
+
|
| 24 |
T2V-Turbo learns a LoRA on top of the base model by aligning to the reward feedback from [HPSv2.1](https://github.com/tgxs002/HPSv2/tree/master) and [InternVid2 Stage 2 Model](https://huggingface.co/OpenGVLab/InternVideo2-Stage2_1B-224p-f4).
|
| 25 |
+
|
| 26 |
T2V-Turbo-v2 optimizes the training techniques by finetuning the full base model and further aligns to [CLIPScore](https://huggingface.co/laion/CLIP-ViT-H-14-laion2B-s32B-b79K)
|
| 27 |
|
| 28 |
T2V-Turbo trains on pure WebVid-10M data, whereas T2V-Turbo-v2 carufully optimizes different learning objectives with a mixutre of VidGen-1M and WebVid-10M data.
|
| 29 |
|
| 30 |
Moreover, T2V-Turbo-v2 supports to distill motion priors from the training videos.
|
| 31 |
|
| 32 |
+
[Project page for T2V-Turbo](https://t2v-turbo.github.io) 🥳
|
| 33 |
+
|
| 34 |
+
[Project page for T2V-Turbo-v2](https://t2v-turbo-v2.github.io) 🤓
|
| 35 |
"""
|
| 36 |
if torch.cuda.is_available():
|
| 37 |
DESCRIPTION += "\n<p>Running on CUDA 😀</p>"
|