---
license: apache-2.0
language:
- en
base_model:
- mistralai/Mistral-7B-Instruct-v0.3
---
model_name: "Mistral-7B-Math (LoRA Adapters Only)" \
repo: "samzheng/mistral-7b-math-lora"\
description: 
  This repo contains only the LoRA adapter weights (~200 MB) trained for 
  grade-school symbolic math. Load them on top of the 4-bit base model 
  to save disk and download time.

quick_start:

    from transformers import AutoModelForCausalLM, AutoTokenizer
    from peft import PeftModel

    base_id = "unsloth/mistral-7b-instruct-v0.3-bnb-4bit"
    lora_id = "yourusername/mistral-7b-math-lora"

    tok   = AutoTokenizer.from_pretrained(base_id)
    base  = AutoModelForCausalLM.from_pretrained(base_id, load_in_4bit=True, device_map="auto")
    model = PeftModel.from_pretrained(base, lora_id)   # inject adapters

    # generate
    prompt = "... Alpaca-formatted prompt ..."
    out = model.generate(**tok(prompt, return_tensors="pt").to(model.device), max_new_tokens=256)
    print(tok.decode(out[0], skip_special_tokens=True))