metadata
library_name: transformers
license: apache-2.0
base_model: ibm-granite/granite-3.0-1b-a400m-base
tags:
- axolotl
- moe
- roleplay
model-index:
- name: MoE_Girl_400MA_1BT
results: []
QuantFactory/MoE-Girl_400MA_1BT-GGUF
This is quantized version of allura-org/MoE-Girl_400MA_1BT created using llama.cpp
Original Model Card
MoE Girl 400mA 1bT
a finetune of Granite 3.0 by IBM designed for roleplaying (and maybe general usecases if you try hard enough).
Disclaimer
PLEASE do not expect godliness out of this, it's a model with 400 million active parameters. Expect something more akin to GPT-2.
Quants
TODO!
Prompting
Use ChatML.
<|im_start|>system
You are a helpful assistant who talks like a pirate.<|im_end|>
<|im_start|>user
Hello there!<|im_end|>
<|im_start|>assistant
Yarr harr harr, me matey!<|im_end|>
Thanks
Special thanks to the members of Allura for testing and emotional support, as well as the creators of all the datasets that were used in the Special Sauce used to train this model. I love you all <3 - Fizz