base_model: | |
- ibm-granite/granite-3.1-8b-instruct | |
In testing. If this doesn't suck I'll write a full model card. | |
Granite instruct or Fizzpaca (Alpaca with System/Instruction/Response) | |
Working samplers so far: | |
- temp 0.8, top-a 0.3 | |
- temp 0.6, min-p 0.1 | |