LLaVA-NeXT
LLaVA-NeXT copied to clipboard
LLaVA-VL
https://github.com/LLaVA-VL/LLaVA-NeXT/blob/inference/docs/LLaVA-NeXT.md In this example, your code generate double "" in front of "user" for the prompt_question variable. Could you check if there is any mistake in your code. Below is...
As describeed in blog-2024-05-25(https://llava-vl.github.io/blog/2024-05-25-llava-next-ablations/), higher-AnyRes is proposed to avoid the loss of detail for high-resolution images. Where can I find the **code of** **higher-AnyRes image dividing method** and the **thresholded...
code：video = image_processor.preprocess(video, return_tensors="pt")["pixel_values"].half().cuda()
Hello, I am trying to find the training code, but it seems like there is just inference code. Can you please point to the training code?
Gradio is fine for playing around but can you please add proper demo code like you've done for the other llava-next models?
Hi, team. Thank you for your great work. I made some ad-hoc changes to use the interleave model. Please let me know if I need to change something.
`bash playground/demo/interleave_demo.py --model_path path/to/ckpt` The execute code should be run with python not bash. And How can this code specify the input image sequence? It appears to be just a...
Hello LLaVa-NeXT team! I want to clarify some points about the AnyRes technique and how the image feature is unpadded in modeling forward. As this [issue](https://github.com/huggingface/transformers/issues/31327) shows, seems like a...
Hi, Dear author: It seems the llava-next is really insightful exploreing work. Please kindly release the training and inference code asap, thank you very much.
