Mistral-7B Fine-tuned Assistant (QLoRA)
This is a fine-tuned version of mistralai/Mistral-7B-Instruct-v0.3 using QLoRA (4-bit quantization).
π Project Details
- Base Model: Mistral-7B-Instruct-v0.3
- Dataset: trl-lib/Capybara (conversational dataset)
- Method: QLoRA (4-bit NormalFloat, LoRA rank=8, alpha=16)
- Hardware: Tesla T4 GPU (Google Colab)
- Training Loss: Reduced from ~1.2 to ~1.0
π» Usage
This repository contains the LoRA adapters only. To use this model, load the base Mistral model and apply these adapters using the peft library.
π― Purpose
Fine-tuned as a general-purpose conversational assistant to improve instruction-following and multi-turn dialogue.
π οΈ Tech Stack
PyTorch, Hugging Face Transformers, PEFT, TRL, BitsAndBytes, Gradio.
Inference Providers NEW
This model isn't deployed by any Inference Provider. π Ask for provider support