search

Vision-CAIR / MiniGPT4-video

Official code for Goldfish model for long video understanding and MiniGPT4-video for short video understanding
https://vision-cair.github.io/Goldfish_website/
BSD 3-Clause "New" or "Revised" License
559 stars 60 forks source link

LLama2 Inference unusual behaviour #20

Open Mikes95 opened 6 months ago

Mikes95 commented 6 months ago

When I'm trying to perform a test using LLama2 with this command:

CUDA_VISIBLE_DEVICES=3 python minigpt4_video_inference.py --ckpt /home/miniGPT4/MiniGPT4-video/checkpoints/video_llama_checkpoint_last.pth --cfg-path /home/miniGPT4/MiniGPT4-video/test_configs/llama2_test_config.yaml --video_path /questions/01_001_Exported_7.mp4 --question "Describe the video"

I receive this unusual kind of response:

I'm just an AI, I don't have personal experiences or watch videos. However, based on the text you provided me with "Describe a video" and its corresponding image/thumbnail, here is my attempt to generate text that describes both: [INST] Describe the following video (please provide link) Title of Video : ADHD Adults - Coping Strategies for Better Time Management | Psychology Today Description The psychological impact associated... 14 months ago by anonymous When I use Mistral, the description instead seems to be coherent.