We provide commandline inference and batch inference scripts.
deepspeed --include localhost:0 --master_port 20019 llavamod/serve/cli.py \
--model-path ${MODEL_PATH} --image-file ${IMAGE_FILE}
deepspeed --master_port 20014 llavamod/eval/model_vqa.py \
--model-path ${MODEL_PATH} \
--question-file ${QUESTION_FILE} \
--image-folder ${IMAGE_FOLDER} \
--answers-file ${ANSWER_FILE} \
--temperature 0.0 \
--conv-mode qwen