From 67967e3c7296737f09f43acdcfee7e42560d7be7 Mon Sep 17 00:00:00 2001 From: Hyunsu Philip Cho Date: Tue, 16 Aug 2022 03:42:26 -0700 Subject: [PATCH] Use G4 instances to avoid capacity issue --- aws/rapids_studio_hpo.ipynb | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/aws/rapids_studio_hpo.ipynb b/aws/rapids_studio_hpo.ipynb index 23ff4b29..9ca3c767 100644 --- a/aws/rapids_studio_hpo.ipynb +++ b/aws/rapids_studio_hpo.ipynb @@ -196,7 +196,7 @@ "metadata": {}, "source": [ "Depending on the workflow you have chosen, your instance should reflect the specifications needed. For example, for the singleGPU workflow, you should choose an instance with a GPU, such as the p3.2xlarge instance. You can [read about Amazon EC2 Instance Types here](https://aws.amazon.com/ec2/instance-types/). \n", - "> e.g., For the 10_year dataset option, we suggest ml.p3.8xlarge instances (4 GPUs) and ml.m5.24xlarge CPU instances ( we will need upwards of 200GB CPU RAM during model training)." + "> e.g., For the 10_year dataset option, we suggest ml.g4dn.12xlarge instances (4 GPUs) and ml.m5.24xlarge CPU instances ( we will need upwards of 200GB CPU RAM during model training)." ] }, { @@ -208,7 +208,7 @@ "outputs": [], "source": [ "# we will recommend a compute instance type, feel free to modify \n", - "instance_type = 'ml.p3.2xlarge' # recommend_instance_type(ml_workflow_choice, dataset_directory)" + "instance_type = 'ml.g4dn.2xlarge' # recommend_instance_type(ml_workflow_choice, dataset_directory)" ] }, {