File size: 892 Bytes
beefb8e 0b66d3a 5ff2646 beefb8e 5ff2646 beefb8e |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 |
# AskVideos-7B-Instruct-v0.1
## Model details
**Model type:**
AskVideos-7B-Instruct-v0.1 is an open-source chatbot trained by fine-tuning a Video-LLaMA variant on additional video Q&A data.
It uses a frozen Vicuna 7B v1.1 LLM to answer Video-Text queries and a frozen BLIP style image encoder.
A video feature is derived from the encoded image using a video-QFormer and the result is projected onto the LLM space.
**Github repo for demo:**
https://github.com/AskYoutubeAI/AskVideos
**Acknowledgement**
This model is based on Video-LLaMA. Check out the original work here: https://github.com/DAMO-NLP-SG/Video-LLaMA
## License
AskVideos-7B-Instruct-v0.1 code and models are distributed under the Apache License 2.0.
## Training dataset
- 50K video synthetic Q&A pairs mined from videos.
- Trained with 16 images sampled over 30s clips per Q&A pair.
- Finetuned on Video-LLaAMA Vicuna 7B. |