LoRA Fine-Tuned LLM (Colab GPU → Local Inference)
Автор: MVENDAGA NGUAKAAGA
Загружено: 2025-12-18
Просмотров: 38
Описание:
Project 9 — LoRA Fine-Tuned LLM (Colab GPU → Local Inference)
This video demonstrates end-to-end LoRA fine-tuning of a small open-source language model and local inference using the saved adapters.
What’s shown in this demo
Instruction–response dataset used for training
LoRA fine-tuning on Google Colab (T4 GPU)
Parameter-efficient training using PEFT (LoRA)
Saving LoRA adapter weights
Running local inference on CPU
Before vs after comparison (base model vs LoRA-fine-tuned model)
Tech Stack
Python
HuggingFace Transformers
Datasets
PEFT (LoRA)
Torch
Google Colab (GPU training)
Local inference (no cloud APIs)
Base Model
TinyLlama/TinyLlama-1.1B-Chat-v1.0
Why this project matters
This project proves the ability to:
Fine-tune LLMs efficiently without retraining full models
Work with GPU training environments (Colab)
Package and reuse LoRA adapters
Run inference locally without paid APIs
Build reproducible, portfolio-ready LLM systems
Repository
📦 GitHub: link in README
Demo Type
Local run (developer mode)
No OpenAI API
No paid hosting
Tags (optional but recommended)
LLM Engineering
LoRA Fine Tuning
PEFT
HuggingFace
Transformers
TinyLlama
AI Engineering Portfolio
Local LLM
Colab GPU
Повторяем попытку...
Доступные форматы для скачивания:
Скачать видео
-
Информация по загрузке: