The result of fine tuning llama-8b by mistake when I thought I had created a frankenmerge. Testing suggests it works alright though and with the Alpaca prompt format.
Trained for 3 epochs on the AdamCodd/no_robots-alpaca dataset with V100.
Uploaded model
- Developed by: reallad
- License: apache-2.0
- Finetuned from model : reallad/frankenllama-16b-untuned
This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.
- Downloads last month
- 9
Inference Providers
NEW
This model is not currently available via any of the supported Inference Providers.