feifeiobama commited on
Commit
cb3e3c1
1 Parent(s): f67483e

Add paper link

Browse files
Files changed (1) hide show
  1. README.md +11 -4
README.md CHANGED
@@ -6,7 +6,7 @@ base_model:
6
 
7
  # ⚡️Pyramid Flow⚡️
8
 
9
- [[Paper]](https://arxiv.org) [[Project Page ✨]](https://pyramid-flow.github.io) [[Code ⚡️]](https://github.com/jy0205/Pyramid-Flow)
10
 
11
  This is the official repository for Pyramid Flow, a training-efficient **Autoregressive Video Generation** method based on **Flow Matching**. By training only on open-source datasets, it generates high-quality 10-second videos at 768p resolution and 24 FPS, and naturally supports image-to-video generation.
12
 
@@ -26,12 +26,19 @@ This is the official repository for Pyramid Flow, a training-efficient **Autoreg
26
  ## News
27
 
28
  * `COMING SOON` ⚡️⚡️⚡️ Training code and new model checkpoints trained from scratch.
29
- * `2024.10.10` 🚀🚀🚀 We release the [technical report](https://arxiv.org), [project page](https://pyramid-flow.github.io) and [model checkpoint](https://huggingface.co/rain1011/pyramid-flow-sd3) of Pyramid Flow.
30
 
31
  ## Usage
32
 
33
  You can directly download the model from [Huggingface](https://huggingface.co/rain1011/pyramid-flow-sd3). We provide both model checkpoints for 768p and 384p video generation. The 384p checkpoint supports 5-second video generation at 24FPS, while the 768p checkpoint supports up to 10-second video generation at 24FPS.
34
 
 
 
 
 
 
 
 
35
  To use our model, please follow the inference code in `video_generation_demo.ipynb` at [this link](https://github.com/jy0205/Pyramid-Flow/blob/main/video_generation_demo.ipynb). We further simplify it into the following two-step procedure. First, load the downloaded model:
36
 
37
  ```python
@@ -44,7 +51,7 @@ torch.cuda.set_device(0)
44
  model_dtype, torch_dtype = 'bf16', torch.bfloat16 # Use bf16, fp16 or fp32
45
 
46
  model = PyramidDiTForVideoGeneration(
47
- '/home/jinyang06/models/pyramid-flow', # The downloaded checkpoint dir
48
  model_dtype,
49
  model_variant='diffusion_transformer_768p', # 'diffusion_transformer_384p'
50
  )
@@ -133,7 +140,7 @@ Consider giving this repository a star and cite Pyramid Flow in your publication
133
  @article{jin2024pyramidal,
134
  title={Pyramidal Flow Matching for Efficient Video Generative Modeling},
135
  author={Jin, Yang and Sun, Zhicheng and Li, Ningyuan and Xu, Kun and Xu, Kun and Jiang, Hao and Zhuang, Nan and Huang, Quzhe and Song, Yang and Mu, Yadong and Lin, Zhouchen},
136
- jounal={arXiv preprint arXiv:2410.XXXXX},
137
  year={2024}
138
  }
139
  ```
 
6
 
7
  # ⚡️Pyramid Flow⚡️
8
 
9
+ [[Paper]](https://arxiv.org/abs/2410.05954) [[Project Page ✨]](https://pyramid-flow.github.io) [[Code 🚀]](https://github.com/jy0205/Pyramid-Flow)
10
 
11
  This is the official repository for Pyramid Flow, a training-efficient **Autoregressive Video Generation** method based on **Flow Matching**. By training only on open-source datasets, it generates high-quality 10-second videos at 768p resolution and 24 FPS, and naturally supports image-to-video generation.
12
 
 
26
  ## News
27
 
28
  * `COMING SOON` ⚡️⚡️⚡️ Training code and new model checkpoints trained from scratch.
29
+ * `2024.10.10` 🚀🚀🚀 We release the [technical report](https://arxiv.org/abs/2410.05954), [project page](https://pyramid-flow.github.io) and [model checkpoint](https://huggingface.co/rain1011/pyramid-flow-sd3) of Pyramid Flow.
30
 
31
  ## Usage
32
 
33
  You can directly download the model from [Huggingface](https://huggingface.co/rain1011/pyramid-flow-sd3). We provide both model checkpoints for 768p and 384p video generation. The 384p checkpoint supports 5-second video generation at 24FPS, while the 768p checkpoint supports up to 10-second video generation at 24FPS.
34
 
35
+ ```python
36
+ from huggingface_hub import snapshot_download
37
+
38
+ model_path = 'PATH' # The local directory to save downloaded checkpoint
39
+ snapshot_download("rain1011/pyramid-flow-sd3", local_dir=model_path, local_dir_use_symlinks=False, repo_type='model')
40
+ ```
41
+
42
  To use our model, please follow the inference code in `video_generation_demo.ipynb` at [this link](https://github.com/jy0205/Pyramid-Flow/blob/main/video_generation_demo.ipynb). We further simplify it into the following two-step procedure. First, load the downloaded model:
43
 
44
  ```python
 
51
  model_dtype, torch_dtype = 'bf16', torch.bfloat16 # Use bf16, fp16 or fp32
52
 
53
  model = PyramidDiTForVideoGeneration(
54
+ 'PATH', # The downloaded checkpoint dir
55
  model_dtype,
56
  model_variant='diffusion_transformer_768p', # 'diffusion_transformer_384p'
57
  )
 
140
  @article{jin2024pyramidal,
141
  title={Pyramidal Flow Matching for Efficient Video Generative Modeling},
142
  author={Jin, Yang and Sun, Zhicheng and Li, Ningyuan and Xu, Kun and Xu, Kun and Jiang, Hao and Zhuang, Nan and Huang, Quzhe and Song, Yang and Mu, Yadong and Lin, Zhouchen},
143
+ jounal={arXiv preprint arXiv:2410.05954},
144
  year={2024}
145
  }
146
  ```