Skip to content

The GSM8K evaluation result of Llama-3-8B-instruct model is only 0.44 #26

Description

@Vance0124

Are there any important considerations when evaluating the Llama-3-8B-instruct model? My evaluation result is only 0.44.
I use the command:

python eval_math.py --model /xxx/pretrain/Meta-Llama-3-8B-Instruct --data_file /xxx/eval_math/GSM8K_test_data.jsonl --save_path /xxx/Meta-Llama-3-8B-Instruct.json --tensor_parallel_size 8 --seed 42

image

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions