search

mlpc-ucsd / BLIVA

(AAAI 2024) BLIVA: A Simple Multimodal LLM for Better Handling of Text-rich Visual Questions
https://arxiv.org/abs/2308.09936
BSD 3-Clause "New" or "Revised" License
259 stars 26 forks source link

The evaluate quesion:when load_state_dict #26

Open mingtouyizu opened 1 month ago

mingtouyizu commented 1 month ago

When I use the evaluation code of BLIVA, everything is normal. However, when I use the evaluation code of LAVIS, there is an error of missing model parameters: The model used is as follows: ... . The location of the error in the code is as follows: image Missing parameter picture: image

mingtouyizu commented 1 month ago

Can anyone sovle my issue?

mingtouyizu commented 1 month ago

I have identified the issue; it occurred when running evaluate.py with torch.distributed.run --nproc_per_node=4. After removing it, the program runs normally. But what I want to know now is how to use torch.distributed.run.