metadata
base_model: mlx-community/Mistral-7B-Instruct-v0.3-4bit
library_name: mlx
tags:
- lora
- mlx
- finance
bha6kar/finance-lora-mistral-7b
LoRA adapter fine-tuning mlx-community/Mistral-7B-Instruct-v0.3-4bit on the gbharti/finance-alpaca financial instruction dataset with MLX.
- Test perplexity: 12.078 (base) to 4.648 (tuned), across 3 seed(s).
Use
from mlx_lm import load, generate
model, tok = load("mlx-community/Mistral-7B-Instruct-v0.3-4bit", adapter_path="<downloaded-adapter-dir>")
Produced by slm-training.