Examples of fine-tuning LLMs and deployment using Azure ML distributed compute (Multiple GPUs & Multiple nodes) Fine-tuning help you improve model's quality and consistency in specialized scenerios. This repo fine-tunes pretrained models (LLAMA2-7B, LLAMA2_13B or LLAMA2-70B, including Chat models) from Azure ML's model registry using Hugging Face's SFT library. Azure ML distributed DL infrastructure allow easy scaling out for large scale training. The fine-tuned model is registered in MLFLow format to Azure ML.
Instruction:
- For fine-tuning:
-
Setup your Azure ML CLI v2: https://learn.microsoft.com/en-us/azure/machine-learning/how-to-configure-cli?view=azureml-api-2&tabs=public
-
Make sure you have A100 GPU SKU (NCadsA100 or NDadsA100) series
-
Checkout
finetine_pipeline.ymlfor training parameters and update. Note some important parameters such as whether you're fine-tuning a Chat model vs. regular model because the format of the prompt is different in Chat model. -
Go to llma2 folder Run the training script:
az ml job create finetune_pipeline.yml -
Use the test.ipynb notebook to test the fine-tuned model.
- For deployment
- Create online endpoint:
az ml online-endpoint create -f deployment/endpoint.yml - Create the deployment:
az ml online-deployment create -f deployment/deployment.yml - Use the sample in test.ipynb to test the online endpoint Credit: This repo uses training data from from https://github.com/tatsu-lab/stanford_alpaca/tree/main