How thirumurugan240/fine_tuning_concepts works
This repository is a hands-on fine-tuning course, with notebooks teaching the workflow from model selection and data preparation through training, evaluation, inference, and serving. The intended artifact path is an adapter produced by notebook 11, used in notebooks 13–14, then merged for notebooks 15–16 and the FastAPI app. Notebook capabilities are inferred from filenames; the notebooks themselves were not excerpted, so other internal calls and transitions are not shown.
Main entry points: merged_model Merged Model · app.py Generation API · finetuned_model LoRA Adapter · 14_Merging_LoRA_and_Saving.ipynb Adapter Merging
Components 20
Fine-Tuning Foundations
- Fine-Tuning Intro
01_Introduction_to_Fine_Tuning.ipynb - Method Comparison
02_FineTuning_vs_Prompting_vs_RAG.ipynb - Training Pipeline
03_LLM_Training_Pipeline.ipynb - Fine-Tuning Types
04_Types_of_Fine_Tuning.ipynb - Hardware Setup
05_Hardware_and_Environment_Setup.ipynb - Base Model Choice
06_Choosing_the_Base_Model.ipynb
Data Preparation
- Dataset Concepts
07_Understanding_Datasets.ipynb - Dataset Creation
08_Create_Your_Own_Dataset.ipynb - Tokenization Setup
09_Tokenization_and_Chat_Templates.ipynb
Training and Evaluation
- LoRA Configuration
10_LoRA_QLoRA_Configuration.ipynb - Model Fine-Tuning
11_Fine_Tuning_Your_First_LLM.ipynb - LoRA Adapter
finetuned_model - Training Monitoring
12_Monitoring_and_Overfitting.ipynb - Model Evaluation
13_Evaluating_the_Model.ipynb - Adapter Merging
14_Merging_LoRA_and_Saving.ipynb
Inference and Serving
- Merged Model
merged_model - Model Inference
15_Running_Inference.ipynb - API Tutorial
16_Building_an_API_with_FastAPI.ipynb - Generation API
app.py
Other components
- API Caller
Connections 9
- Model Fine-Tuning → LoRA Adapter: creates
- Model Evaluation → LoRA Adapter: uses
- Adapter Merging → LoRA Adapter: uses
- Adapter Merging → Merged Model: creates
- Model Inference → Merged Model: uses
- API Tutorial → Merged Model: uses
- Generation API → Merged Model: loads
app.py - API Caller → Generation API: requests generation
app.py - Generation API → API Caller: returns answer
app.py