Video-Coder training data
Browse files- .gitattributes +1 -0
- README.md +58 -0
- cached_image/cached_image_part1.zip +3 -0
- gt_video/gt_video.zip +3 -0
- rl_data.json +0 -0
- sft_data.json +3 -0
.gitattributes
CHANGED
|
@@ -58,3 +58,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 58 |
# Video files - compressed
|
| 59 |
*.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 60 |
*.webm filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 58 |
# Video files - compressed
|
| 59 |
*.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 60 |
*.webm filter=lfs diff=lfs merge=lfs -text
|
| 61 |
+
sft_data.json filter=lfs diff=lfs merge=lfs -text
|
README.md
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
---
|
| 4 |
+
|
| 5 |
+
|
| 6 |
+
# Video-Coder Training Data
|
| 7 |
+
|
| 8 |
+
|
| 9 |
+
This repository contains the training data for [Video-Coder: Reinforcing Agentic Coding for Video Generation]().
|
| 10 |
+
|
| 11 |
+
|
| 12 |
+
[**๐ Project Page**]() | [**๐ Paper**]() | [**๐ป Code**]()
|
| 13 |
+
|
| 14 |
+
|
| 15 |
+
# ๐ Intro
|
| 16 |
+
|
| 17 |
+
|
| 18 |
+
<div align="center">
|
| 19 |
+
<img src="https://raw.githubusercontent.com/CYWang735/Video-Coder/main/assets/teaser.png" alt="Video-Coder Overview" width="80%">
|
| 20 |
+
</div>
|
| 21 |
+
|
| 22 |
+
|
| 23 |
+
**Video-Coder** is a code-driven agent for video generation. It iteratively writes, executes, and refines Blender Python programs to construct dynamic 3D scenes and render draft videos, providing structural and spatial guidance for downstream video generators.
|
| 24 |
+
|
| 25 |
+
|
| 26 |
+
We design a dedicated data pipeline and curate two high-quality datasets, **Video-Coder-SFT-10k** and **Video-Coder-RL-4k**, which contain diverse prompts and ground-truth reference videos.
|
| 27 |
+
|
| 28 |
+
|
| 29 |
+
## Dataset Files
|
| 30 |
+
|
| 31 |
+
|
| 32 |
+
The Video-Coder training data consists of JSON files tailored for different training stages:
|
| 33 |
+
* `sft_data.json`: Used for Supervised Fine-Tuning (SFT) cold start.
|
| 34 |
+
* `rl_data.json`: Used for Reinforcement Learning (RL) training using GRPO.
|
| 35 |
+
|
| 36 |
+
|
| 37 |
+
## ๐ฅ Demo
|
| 38 |
+
|
| 39 |
+
|
| 40 |
+
|
| 41 |
+
|
| 42 |
+
<div align="center">
|
| 43 |
+
<img src="https://raw.githubusercontent.com/CYWang735/Video-Coder/main/assets/qualitative.png" alt="Inference Process" width="85%">
|
| 44 |
+
</div>
|
| 45 |
+
|
| 46 |
+
|
| 47 |
+
|
| 48 |
+
|
| 49 |
+
## Citation
|
| 50 |
+
|
| 51 |
+
|
| 52 |
+
If you find this work helpful for your research, please consider citing:
|
| 53 |
+
|
| 54 |
+
|
| 55 |
+
```bibtex
|
| 56 |
+
|
| 57 |
+
|
| 58 |
+
```
|
cached_image/cached_image_part1.zip
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6ce0335a20f846e7a63e7fbb4626f02e9cc2475734c83d23c9f8ce26153a9298
|
| 3 |
+
size 1421777050
|
gt_video/gt_video.zip
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:af1912c9e19b602589846675f9ac599fc35d9a0ff13936556115f215d3c66e1e
|
| 3 |
+
size 581792571
|
rl_data.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
sft_data.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bb35dc15fc04a9b859d665cdff793fb645952a5886298db83a8f39ac1ba2c259
|
| 3 |
+
size 723624929
|