LLaVA-NeXT

LLaVA-NeXT copied to clipboard

What is the difference between the two projects, lmms-lab/llava-onevision-qwen2-7b-ov and lmms-lab/llava-onevision-qwen2-7b-si?

1

I looked at the model card introduction but didn't see what the main differences are between these two models. Could the author explain?

AmazDeng

if `image_idx not in video_idx_in_batch`, `image_feature` will be added into `new_image_features` repeatedly, which should be avoided. That's what this PR does.

Wang-Xiaodong1899

The llava-onevision model video inference code has an error

16

For the llava-onevision model, the official video inference code does not modify the `image_aspect_ratio` parameter, resulting in the use of the default `anyres_max_9`. This causes the `image_features` to occupy a...

AmazDeng

https://discord.gg/ek27DUnYdS

dandan-three

For the first version of the llava-next-video project, the model chosen was LLaVA-NeXT-Video-7B-DPO. If the number of frames is set to 32, the final inputs_embeds dimension sent to the llama2...

AmazDeng

Where is the code of Bilinear Interpolation?

1

As title

liuao743

Missing mm_projector in latest LLaVa

11

I am attempting to run the `finetune_onevision.sh` script. I've gotten many things sorted out but I am stumped by the `--pretrain_mm_mlp_adapter` argument. The default value as provided in the script...

mrd

interleave_demo.py missing

2

Hi, thank you very much for you research, it is very interesting! I am interested in running the LLaVA-Next Interleave model but the file playground/demo/interleave_demo.py is missing Can i find...

ale93111

The checkpoint of LLaVA-NeXT-Video stage1.

2

Hi! I wonder know whether you have the plan to release the checkpoint of LLaVA-NeXT-Video stage 1, the pretrain version. I want to finetune the model from this stage!

zhengrongz

Weight size mismatch when load_pretrain for llava_onevision (0.5b model)

4

Hi~,I am recently trying to use the llava_onevision model, I try to follow the onevision tutorial, which seems pretty easy. I run the program exactly as the tutorial, the model...

zeal-up