Using the qwen 2.5 math test script to reproduce the results of Qwen2.5-14B-Instruct failed

[SefaZeng]

This is the scripts:

TOKENIZERS_PARALLELISM=false \
python3 -u ${WORK_PATH}/math_eval.py \
    --model_name_or_path ${MODEL_NAME_OR_PATH} \
    --data_name ${DATA_NAME} \
    --data_dir ${WORK_PATH}/data \
    --output_dir ${OUTPUT_DIR} \
    --split ${SPLIT} \
    --prompt_type ${PROMPT_TYPE} \
    --num_test_sample ${NUM_TEST_SAMPLE} \
    --max_tokens_per_call 16384 \
    --seed 0 \
    --temperature 0 \
    --n_sampling 1 \
    --top_p 1 \
    --top_k 1 \
    --start 0 \
    --end -1 \
    --save_outputs \
    --overwrite \
    --use_vllm \
    --apply_chat_template

[jklj077]

ping @hzhwcmhf , who knows evaluation better

[hzhwcmhf]

@SefaZeng Can you provide the model outputs?