File size: 467 Bytes
b0c0df0 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 |
# pip3 install vllm
# pip3 install qwen_vl_utils
# cd ~/prod/lmms-eval-public
# pip3 install -e .
export NCCL_BLOCKING_WAIT=1
export NCCL_TIMEOUT=18000000
export NCCL_DEBUG=DEBUG
python3 -m lmms_eval \
--model vllm \
--model_args model=Qwen/Qwen2-VL-7B-Instruct,tensor_parallel_size=4 \
--tasks mme,gsm8k_cot_self_consistency,mmmu_val \
--batch_size 64 \
--log_samples \
--log_samples_suffix vllm \
--output_path ./logs \
--limit=64 |