10 lines
985 B
Markdown
10 lines
985 B
Markdown
# OpenPipe demo: fine-tuning your own model
|
|
|
|
Hi there! This repository should give you a brief overview of how to fine-tune a competitive model from start to finish. You should review the notebooks in this directory in the following order:
|
|
|
|
1. [./generate-data.ipynb](./generate-data.ipynb): Demonstrates how to generate a sample dataset of GPT-4 completions, store it using OpenPipe, and then export it in a format suitable for training a model.
|
|
2. [./train.ipynb](./train.ipynb): Trains a Llama 2 7B model on the dataset from step (1).
|
|
3. [./evaluate.ipynb](./evaluate.ipynb): Evaluates the model we trained using a special test set that we set aside in step (1).
|
|
4. [./benchmark.ipynb](./benchmark.ipynb): A script to compare costs and completion latencies between our fine-tuned model, GPT-3.5, and GPT-4.
|
|
|
|
If you want to follow along yourself, I recommend using [RunPod](https://www.runpod.io/). The training scripts we use will run on any of their GPUs with 24GB of vRAM or more. |