The model uses 200k samples from different datasets to fine-tune Open-Orca/Mistral-7B-OpenOrca.
| Humaneval | Java | Javascript | tokens/s | |
|---|---|---|---|---|
| Phi-1 | 51.22 | 10.76 | 19.25 | |
| CodeLlama-34b-Instruct | 50.79 | 41.53 | 45.85 | 15.1 |
| CodeLlama-13b-Instruct | 50.6 | 33.99 | 40.92 | 25.3 |
| Speechless Code 7B | 46.34 | |||
| CodeLlama-7b-Instruct | 45.65 |
Its coding performance in Python is comparable to Phi-1.5B.
You must log in or register to comment.