[WIP]Add packer support for building AWS AMI

.github/workflows/build-ami.yml

@@ -0,0 +1,78 @@
+name: Build AWS Neuron AMI
+
+on:
+    push:
+      branches:
+        - main
+      paths:
+        - 'infrastructure/ami/**'
+    pull_request:
+      branches:
+        - main
+      paths:
+        - 'infrastructure/ami/**'
+    workflow_dispatch:
+      inputs: 
+        tag: 
+          description: 'Tag to use for the AMI build'
+          default: 'main'
+    schedule:
+      # Schedule the workflow to run every second day at midnight UTC
+      - cron: '0 0 */2 * *'
+
+jobs:
+    build-ami:
+        defaults:
+          run:
+            working-directory: infrastructure/ami
+        runs-on: ubuntu-latest
+        env:
+          AWS_REGION: us-east-1
+        steps:
+          - name: Checkout
+            uses: actions/checkout@v3
+            with: 
+              ref: ${{ github.event.inputs.tag || github.ref }}
+
+          - name: Setup Packer
+            uses: hashicorp/setup-packer@main
+
+          - name: configure aws credentials
+            uses: aws-actions/configure-aws-credentials@v1
+            with:
+              aws-access-key-id: ${{ secrets.AWS_ACCESS_KEY_ID_BUILD_AMI }}
+              aws-secret-access-key: ${{ secrets.AWS_ACCESS_KEY_SECRET_BUILD_AMI }}
+              aws-region: ${{ env.AWS_REGION }}
+
+          - name: Packer format
+            id: format
+            run: packer fmt hcl2-files
+            continue-on-error: true
+
+          - name: Packer Init
+            id: init
+            run: packer init hcl2-files
+            continue-on-error: true
+
+          - name: Packer Validate
+            id: validate
+            run: packer validate -var "optimum_version=${{ github.event.inputs.tag || github.event.repository.default_branch }}" -var "region=${{ env.AWS_REGION }}" hcl2-files 
+            continue-on-error: true
+
+          - name: Packer Build
+            id: build
+            run:  | 
+              packer build -var "optimum_version=${{ github.event.inputs.tag || github.event.repository.default_branch }}" -var "region=${{ env.AWS_REGION }}" hcl2-files
+
+          - name: Slack Notification on Failure
+            id: slack
+            uses: slackapi/slack-github-action@v1.25.0
+            if: ${{ failure() && github.event_name == 'schedule' }}
+            with:
+              channel-id: 'C06GAEQJLNN' #copied from slack channel
+              payload: |
+                {
+                  "text": "GitHub Action HuggingFace Neuron AMI Build result: ${{job.status}}"
+                }
+            env:
+              SLACK_BOT_TOKEN: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}

infrastructure/ami/README.md

@@ -0,0 +1,78 @@
+# Building AMI with Packer
+
+This directory contains the files for building AMI using [Packer](https://github.com/hashicorp/packer) that is later published as a AWS Marketplace asset.
+
+
+## Folder Structure
+
+- [hcl2-files](./hcl2-files/) - Includes different files which are used by a Packer pipeline to build an AMI. The files are:
+  - [build.pkr.hcl](./hcl2-files/build.pkr.hcl): contains the [build](https://developer.hashicorp.com/packer/docs/templates/hcl_templates/blocks/build) block, defining the builders to start, provisioning them using [provisioner](https://developer.hashicorp.com/packer/docs/templates/hcl_templates/blocks/build/provisioner), and specifying actions to take with the built artifacts using `post-process`.
+  - [variables.pkr.hcl](./hcl2-files/variables.pkr.hcl): contains the [variables](https://developer.hashicorp.com/packer/docs/templates/hcl_templates/blocks/variable) block, defining variables within your Packer configuration.
+  - [sources.pkr.hcl](./hcl2-files/sources.pkr.hcl): contains the [source](https://developer.hashicorp.com/packer/docs/templates/hcl_templates/blocks/source) block, defining reusable builder configuration blocks.
+  - [packer.pkr.hcl](./hcl2-files/packer.pkr.hcl): contains the [packer](https://developer.hashicorp.com/packer/docs/templates/hcl_templates/blocks/packer) block, used to configure some behaviors of Packer itself, such as the minimum required Packer version needed to apply to your configuration.
+- [scripts](./scripts): contains scripts used by [provisioner](https://developer.hashicorp.com/packer/docs/templates/hcl_templates/blocks/build/provisioner) for installing additonal packages/softwares.
+
+
+### Prerequisites
+ - [Packer](https://developer.hashicorp.com/packer/docs/intro): Packer is an open source tool for creating identical machine images for multiple platforms from a single source configuration.
+
+ - AWS Credentials: You need to have AWS credentials configured on your machine. You can configure AWS credentials using [AWS CLI](https://github.com/aws/aws-cli) or by setting environment variables.
+
+ #### Install Packer on Ubuntu/Debian
+ ```bash
+ curl -fsSL https://apt.releases.hashicorp.com/gpg | sudo apt-key add -
+ sudo apt-add-repository "deb [arch=amd64] https://apt.releases.hashicorp.com $(lsb_release -cs) main"
+ sudo apt-get update && sudo apt-get install packer
+ ```
+
+You can also install Packer for other OS from [here](https://developer.hashicorp.com/packer/tutorials/docker-get-started/get-started-install-cli).
+
+#### Configure AWS Credentials
+
+Using Environment Variables:
+```bash
+export AWS_ACCESS_KEY_ID=<access_key>
+export AWS_SECRET_ACCESS_KEY=<secret_key>
+```
+
+Using AWS CLI:
+```bash
+aws configure sso
+```
+
+There are other ways to configure AWS credentials. You can read more about it [here](https://github.com/aws/aws-cli?tab=readme-ov-file#configuration).
+
+### Build AMI
+
+#### Format Packer blocks
+You can format your HCL2 files locally. This command will update your files in place.
+
+Format a single file:
+```bash
+packer fmt build.pkr.hcl 
+```
+
+Format all files in a directory:
+```bash
+packer fmt ./hcl2-files
+```
+
+#### Validate Packer blocks
+You can validate the syntax and configuration of your files locally. This command will return a zero exit status on success, and a non-zero exit status on failure. 
+
+```bash
+packer validate -var 'region=us-west-2' -var 'optimum_version=v0.0.17' ./hcl2-files
+```
+
+#### Run Packer build
+You can run Packer locally. This command will build the AMI and upload it to AWS.
+
+You need to set variables with no default values using `-var` flag. For example:
+```bash
+packer build -var 'region=us-west-2' -var 'optimum_version=v0.0.17' ./hcl2-files
+```
+
+To trigger a github action workflow manually, you can use GitHub CLI:
+```bash
+gh workflow run build-ami.yml -f tag=<tag>
+```

infrastructure/ami/hcl2-files/build.pkr.hcl

@@ -0,0 +1,29 @@
+build {
+  name = "build-hf-dl-neuron"
+  sources = [
+    "source.amazon-ebs.ubuntu"
+  ]
+  provisioner "shell" {
+    script = "scripts/validate-neuron.sh"
+  }
+  provisioner "shell" {
+    script = "scripts/install-huggingface-libraries.sh"
+    environment_vars = [
+      "TRANSFORMERS_VERSION=${var.transformers_version}",
+      "OPTIMUM_VERSION=${var.optimum_version}",
+    ]
+  }
+  provisioner "shell" {
+    inline = ["echo 'source /opt/aws_neuron_venv_pytorch/bin/activate' >> /home/ubuntu/.bashrc"]
+  }
+  provisioner "file" {
+    source      = "scripts/welcome-msg.sh"
+    destination = "/tmp/99-custom-message"
+  }
+  provisioner "shell" {
+    inline = [
+      "sudo mv /tmp/99-custom-message /etc/update-motd.d/",
+      "sudo chmod +x /etc/update-motd.d/99-custom-message",
+    ]
+  }
+}

infrastructure/ami/hcl2-files/packer.pkr.hcl

@@ -0,0 +1,8 @@
+packer {
+  required_plugins {
+    amazon = {
+      version = ">= 1.2.8"
+      source  = "github.com/hashicorp/amazon"
+    }
+  }
+}

infrastructure/ami/hcl2-files/sources.pkr.hcl

@@ -0,0 +1,15 @@
+source "amazon-ebs" "ubuntu" {
+  ami_name      = "huggingface-neuron-{{isotime \"2006-01-02T15-04-05Z\"}}"
+  instance_type = var.instance_type
+  region        = var.region
+  source_ami    = var.source_ami
+  ssh_username  = var.ssh_username
+  launch_block_device_mappings {
+    device_name           = "/dev/sda1"
+    volume_size           = 512
+    volume_type           = "gp2"
+    delete_on_termination = true
+  }
+  ami_users   = var.ami_users
+  ami_regions = var.ami_regions
+}

infrastructure/ami/hcl2-files/variables.pkr.hcl

@@ -0,0 +1,54 @@
+variable "region" {
+  description = "The AWS region"
+  type        = string
+}
+
+variable "instance_type" {
+  default     = "trn1.2xlarge"
+  description = "EC2 machine type for building AMI"
+  type        = string
+}
+
+variable "source_ami" {
+  default     = "ami-0fbea04d7389bcd4e" 
+  description = "Base Image"
+  type        = string
+  /*
+  To get latest value, run the following command:
+  aws ec2 describe-images \
+      --region us-east-1 \
+      --owners amazon \
+      --filters 'Name=name,Values=Deep Learning AMI Neuron PyTorch 1.13 (Ubuntu 20.04) ????????' 'Name=state,Values=available' \
+      --query 'reverse(sort_by(Images, &CreationDate))[:1].ImageId' \
+      --output text
+  */
+}
+
+variable "ssh_username" {
+  default     = "ubuntu"
+  description = "Username to connect to SSH with"
+  type        = string
+}
+
+variable "optimum_version" {
+  description = "Optimum Neuron version to install"
+  type        = string
+}
+
+variable "transformers_version" {
+  default     = "4.36.2"
+  description = "Transformers version to install"
+  type        = string
+}
+
+variable "ami_users" {
+  default     = ["754289655784", "558105141721"]
+  description = "AWS accounts to share AMI with"
+  type        = list(string)
+}
+
+variable "ami_regions" {
+  default     = ["eu-west-1"]
+  description = "AWS regions to share AMI with"
+  type        = list(string)
+}

infrastructure/ami/scripts/install-huggingface-libraries.sh

@@ -0,0 +1,39 @@
+#!/bin/bash
+
+# Activate the neuron virtual environment
+source /opt/aws_neuron_venv_pytorch/bin/activate
+
+echo "Step: install-hugging-face-libraries"
+
+echo "TRANSFORMERS_VERSION: $TRANSFORMERS_VERSION"
+echo "OPTIMUM_VERSION: $OPTIMUM_VERSION"
+
+pip install --upgrade --no-cache-dir \
+    "transformers[sklearn,sentencepiece,vision]==$TRANSFORMERS_VERSION" \
+    "datasets==2.16.1" \
+    "accelerate==0.23.0" \
+    "diffusers==0.25.0" \
+    "evaluate==0.4.1" \
+    "requests==2.31.0" \
+    "notebook==7.0.6" \
+    "markupsafe==2.1.1" \
+    "jinja2==3.1.2" \
+    "attrs==23.1.0"
+
+echo 'export PATH="${HOME}/.local/bin:$PATH"' >> "${HOME}/.bashrc"
+
+echo "Step: install-and-copy-optimum-neuron-examples"
+git clone -b $OPTIMUM_VERSION https://github.com/huggingface/optimum-neuron.git
+
+cd optimum-neuron
+python setup.py install
+cd ..
+
+mkdir /home/ubuntu/huggingface-neuron-samples/ /home/ubuntu/huggingface-neuron-notebooks/
+mv optimum-neuron/examples/* /home/ubuntu/huggingface-neuron-samples/
+mv optimum-neuron/notebooks/* /home/ubuntu/huggingface-neuron-notebooks/
+rm -rf optimum-neuron
+chmod -R 777 /home/ubuntu/huggingface-neuron-samples /home/ubuntu/huggingface-neuron-notebooks
+
+echo "Step: validate-imports-of-huggingface-libraries"
+bash -c 'python -c "import transformers;import datasets;import accelerate;import evaluate;import tensorboard; import torch;"'

infrastructure/ami/scripts/validate-neuron.sh

@@ -0,0 +1,13 @@
+#!/bin/bash
+echo "Step: validate-neuron-devices"
+neuron-ls
+
+# Activate the neuron virtual environment
+source /opt/aws_neuron_venv_pytorch/bin/activate
+
+python -c 'import torch'
+python -c 'import torch_neuronx'
+
+echo "Installing Tensorboard Plugin for Neuron"
+pip install --upgrade --no-cache-dir \
+    "tensorboard-plugin-neuronx"

infrastructure/ami/scripts/welcome-msg.sh

@@ -0,0 +1,11 @@
+#!/bin/bash
+printf "=============================================================================\n"
+printf "       __|  __|_  )\n"
+printf "       _|  (     /   HuggingFace Deep Learning Neuron AMI (Ubuntu 20.04)\n"
+printf "      ___|\___|___|\n"
+printf "=============================================================================\n"
+printf "Welcome to the HuggingFace Deep Learning Neuron AMI (Ubuntu 20.04)\n"
+printf "* Examples:     /home/ubuntu/huggingface-neuron-samples \n"
+printf "* Notebooks:    /home/ubuntu/huggingface-neuron-notebooks \n"
+printf "* Documentation: https://huggingface.co/docs/optimum-neuron/ \n"
+printf "=============================================================================\n"