snowolf / alpaca-on-amazon-sagemaker

This is a sample about how to run stanford_alpaca on Amazon SageMaker, only for demo use.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

alpaca-on-amazon-sagemaker

This is a sample about how to run stanford_alpaca on Amazon SageMaker, include finetune and inference, this sample is only for demo use.

We use s5cmd, a fast tool for S3 download and upload in parallel, to download pretrained model files from S3 and upload fine-tuned model to S3.

This sample code is tested on Amazon SageMaker by creating a training job with p4d.24xlarge (8 * A100, 40GB), you could run this with p4de.24xlarge (8 * A100, 80GB) of course.

As standford_alpaca update quickly, we will keep update this repo as well.

About

This is a sample about how to run stanford_alpaca on Amazon SageMaker, only for demo use.


Languages

Language:Jupyter Notebook 100.0%