LLaMA-Factory/examples/inference/evaluate.sh

13 lines
316 B
Bash

#!/bin/bash
CUDA_VISIBLE_DEVICES=0 python src/evaluate.py \
--model_name_or_path meta-llama/Llama-2-7b-hf \
--adapter_name_or_path ../../saves/LLaMA2-7B/lora/sft \
--template vanilla \
--finetuning_type lora \
--task mmlu \
--split test \
--lang en \
--n_shot 5 \
--batch_size 4