File size: 1,279 Bytes
b0c0df0 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 |
git clone https://github.com/LLaVA-VL/LLaVA-NeXT
cd LLaVA-NeXT
uv pip install -e .
python3 -m accelerate.commands.launch \
--num_processes=8 \
-m lmms_eval \
--model llava \
--model_args pretrained="liuhaotian/llava-v1.5-7b" \
--tasks mme \
--batch_size 1 \
--log_samples \
--log_samples_suffix llava_v1.5_mme \
--output_path ./logs/
# `conv_template` is an arg of the init function of llava in `lmms_eval/models/llava.py`.
# you could find the corresponding value at LLaVA's code, probably in a dict variable `conv_templates` in `llava/conversations.py`
python3 -m accelerate.commands.launch \
--num_processes=8 \
-m lmms_eval \
--model llava \
--model_args pretrained="liuhaotian/llava-v1.6-mistral-7b,conv_template=mistral_instruct" \
--tasks mme,mmbench_en \
--batch_size 1 \
--log_samples \
--log_samples_suffix llava_v1.5_mme_mmbenchen \
--output_path ./logs/
python3 -m accelerate.commands.launch \
--num_processes=8 \
-m lmms_eval \
--model llava \
--model_args pretrained="liuhaotian/llava-v1.6-34b,conv_template=mistral_direct" \
--tasks mme,mmbench_en \
--batch_size 1 \
--log_samples \
--log_samples_suffix llava_v1.5_mme_mmbenchen \
--output_path ./logs/ |