Text Generation
Transformers
PyTorch
skywork
custom_code

What is the humaneval? and humaneval plus scores?

#2
by rombodawg - opened

You say its trained on code but you havnt posted any coding benchmarks, or even compared it to codellama

How to download the models?

Skywork org

You say its trained on code but you havnt posted any coding benchmarks, or even compared it to codellama

We did not focus on its coding capability. At 2T checkpoint, we did benchmark on Humaneval and MBPP. It scored 17.7 on human eval and 26 on MBPP, respectively.
No further benchmark on code was carried out there after.

weitianwen changed discussion status to closed

Sign up or log in to comment