4,5,8-bit GGUF models for CPU+GPU inference
This model is the one of the moloras (Mixture-of-Multi-LoRAs) experiments.
Extract LoRA modules from below models (all based Mistral-7B-v0.1), each LoRA module has its own unique skills. By using multi-loras, they can be combined together statically or dynamically to form a versatile new model.
- HuggingFaceH4/zephyr-7b-beta (Uncensored Model)
- meetkai/functionary-small-v2.2 (Execute functions/plugins)
- uukuguy/speechless-code-mistral-7b-v1.0 (Enhance Coding)
The entire process is completed through the use of extract-lora, merge-lora, and lora-hub provided by multi-loras.
The router of mixture-of-multi-loras enables an automatic assembling of LoRA modules, using a gradientfree approach to obtain the coefficients of LoRA modules and requiring only a handful of inference steps for unseen tasks.
Code: https://github.com/uukuguy/multi_loras
LM-Evaluation-Harness
| Metric | Value |
|---|---|
| ARC | 61.52 |
| HellaSwag | 83.88 |
| MMLU | 64.71 |
| TruthfulQA | 44.99 |
| Winogrande | 78.69 |
| GSM8K | 43.82 |
| Average | 62.93 |