AWS Builder Center

How to Launch a Trn2 Capacity Block

This guide walks through how to launch a Trn2 instance using Capacity Blocks for ML.

Launching a Trn2 Capacity Block

This guide walks through how to launch a Trn2 instance using Capacity Blocks for ML.

Step 1 — Open the AWS Console Go to  https://us-west-2.console.aws.amazon.com/ 





Step 2 — Select the correct Region

  • We recommend starting with trn2.3xl in Melbourne (ap-southeast-4). If you're working with a model larger than 30B, switch the Region to Ohio (us-east-2).


Step 3 — Open EC2 Search for EC2 in the top search bar and click it.



Step 4 — Go to Capacity Reservations From the EC2 dashboard, click Capacity Reservations in the left-hand menu.



Step 5 — Create a Capacity Reservation Click the orange Create capacity reservation button in the top right.



Step 6 — Choose Capacity Block for ML Select Capacity Block for ML on the right.


Step 7 — Select the instance type Choose trn2.3xl.

Step 8 — Find capacity Click Find capacity.

Step 9 — Pick your capacity block Select the time window (24-hour block) that fits your needs — you can launch same day or next day — then click Create in the bottom right.

Step 10 — Open your reservation Once your capacity reservation is ready, return to the Capacity Reservations tab on the left and click your reservation.

Step 11 — Launch an instance Click the orange Launch instance button in the top right.

Step 12 — Name it and pick an AMI Give the instance a name (e.g., test). Search for neuron


Step 13 — Select the Neuron Deep Learning AMI After typing neuron and pressing search, select one of the Deep Learning AMIs:

Deep Learning AMI Neuron (Amazon Linux 2023), or Deep Learning AMI Neuron (Ubuntu 24.04) — whichever you prefer.

Step 14 — Configure key pair and storage Create your own key pair (you'll need the .pem file to connect). Increase storage to the size you need (e.g., 1000 GB). Click the orange Launch instance button on the right.

Step 15 — Open Instances Click Instances to see your new instance.

Step 16 — Connect Select the instance and click Connect.

Step 17 — SSH in Open the SSH client tab and copy the example SSH command (or hand it to Claude Code / Kiro to help). Run it, and you're in.

Step 18 — Verify Neuron devices Run:
neuron-ls
You should see output like this:
instance-type: trn2.3xlarge
instance-id: i-07c26763c87749b51
logical-neuroncore-config: 2
+--------+--------+----------+--------+--------------+----------+------+
| NEURON | NEURON | NEURON | NEURON | PCI | CPU | NUMA |
| DEVICE | CORES | CORE IDS | MEMORY | BDF | AFFINITY | NODE |
+--------+--------+----------+--------+--------------+----------+------+
| 0 | 4 | 0-3 | 96 GB | 0000:33:00.0 | 0-11 | 0 |
+--------+--------+----------+--------+--------------+----------+------+

That confirms the Neuron devices are visible and the instance is ready to use!
please launch one of our examples !
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article