This repository contains a LoRA (Low-Rank Adaptation) adapter for the Mistral 7B Instruct v0.3 model, fine-tuned to generate high-quality meal reviews. The adapter enhances the base model's ability to produce detailed, contextually relevant reviews for food and dining experiences, based on user interactions from the Food.com dataset.

Model Details

Base Model: mistralai/Mistral-7B-Instruct-v0.3
Fine-Tuning Method: LoRA (Low-Rank Adaptation)
Task: Text generation for meal reviews
Training Data: The Food.com Recipes and User Interactions dataset, specifically the user review text. The dataset contains over 700,000 recipe reviews, which were preprocessed to focus on review generation.
Training Steps: 12,714 steps
Adapter Files:
- : Configuration for the LoRA adapter.
- : Fine-tuned LoRA weights.

Usage

To use this LoRA adapter, merge it with the base Mistral 7B model using the and libraries. Below is an example of how to load and use the adapter for inference.

Installation

pip install transformers peft torch

Example Code

python

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
import torch

# Load base model and tokenizer
base_model_name = "mistralai/Mistral-7B-Instruct-v0.3"
adapter_path = "Oliver1703dk/meal_review_fine_tuned_adapter_bigger"
output_dir = "./merged_model"

tokenizer = AutoTokenizer.from_pretrained(base_model_name)
base_model = AutoModelForCausalLM.from_pretrained(
    base_model_name,
    torch_dtype=torch.float16,
    device_map="auto"
)

# Load LoRA adapter
model = PeftModel.from_pretrained(base_model, adapter_path)

# Merge adapter with base model
merged_model = model.merge_and_unload()

# Save merged model (optional)
merged_model.save_pretrained(output_dir)
tokenizer.save_pretrained(output_dir)

# Inference
prompt = "Write a review for a delicious Italian meal."
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = merged_model.generate(**inputs, max_length=200)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Merged Model

The merged version of this adapter with the base Mistral 7B model is available at Oliver1703dk/meal_reviewstats.io/Oliver1703dk/meal_review_merged_mistral_finetuned_bigger.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Contact

For questions or issues, please open an issue in this repository or contact Oliver1703dk.

Generated on April 29, 2025