ConFusion / scripts /sqa_eval_batch.sh
badayvedat's picture
feat: Add LLaVA model
a824a18
raw
history blame contribute delete
524 Bytes
#!/bin/bash
CHUNKS=8
for IDX in {0..7}; do
CUDA_VISIBLE_DEVICES=$IDX python -m llava.eval.model_vqa_science \
--model-path liuhaotian/llava-lcs558k-scienceqa-vicuna-13b-v1.3 \
--question-file ~/haotian/datasets/ScienceQA/data/scienceqa/llava_test_QCM-LEA.json \
--image-folder ~/haotian/datasets/ScienceQA/data/scienceqa/images/test \
--answers-file ./test_llava-13b-chunk$CHUNKS_$IDX.jsonl \
--num-chunks $CHUNKS \
--chunk-idx $IDX \
--conv-mode llava_v1 &
done