asd / OpenCodeInterpreter /evaluation /evaluate /scripts /04_execution_feedback_multiround_OpenCodeInterpreter.sh
alKoGolik's picture
Upload 210 files
6fcd376 verified
raw
history blame contribute delete
976 Bytes
cd evaluation/evalplus
export PYTHONPATH=$PYTHONPATH:$(pwd)
cd ../evaluate
gpu_start=0
MODEL_NAME_OR_PATHS=(
/ML-A800/home/zhengtianyu/code-interpreter/OpenCodeInterpreter-DS-6.7B
)
DATASET_TYPE="humaneval" #"mbpp"
EVAL_TYPE="plus" #base
for rank in {0..0}; do
gpu_rank=$((rank + gpu_start))
MODEL_NAME_OR_PATH="${MODEL_NAME_OR_PATHS[$rank]}"
echo $MODEL_NAME_OR_PATH
MODEL=$(echo "$MODEL_NAME_OR_PATH" | sed 's|/|_|g')
output_path="./execution_feedback_output"
LOG_DIR="$output_path/$MODEL"
mkdir -p "${LOG_DIR}"
ABS_LOG_DIR=$(realpath "$LOG_DIR")
LOG_FILE="gen_${DATASET_TYPE}_${EVAL_TYPE}_solution_multiround.log"
CUDA_VISIBLE_DEVICES=$gpu_rank nohup python multi_turn/gen_plus_solution_multiround.py \
--model "$MODEL_NAME_OR_PATH" \
--output_path $output_path \
--log_file ${LOG_FILE} \
--version $EVAL_TYPE \
--dataset $DATASET_TYPE \
> "${LOG_DIR}/${LOG_FILE}" 2>&1 &
done