From 8e0c5fe2082763cadd77c4b6b165a60193628343 Mon Sep 17 00:00:00 2001
From: David Corvoysier <david@huggingface.co>
Date: Thu, 27 Jun 2024 11:58:27 +0200
Subject: [PATCH] CI cache builder (#642)

ci: add inference cache builder workflow
---
 .github/workflows/inference_cache.yml | 61 +++++++++++++++++++++++++++
 1 file changed, 61 insertions(+)
 create mode 100644 .github/workflows/inference_cache.yml

diff --git a/.github/workflows/inference_cache.yml b/.github/workflows/inference_cache.yml
new file mode 100644
index 000000000..49dd1565b
--- /dev/null
+++ b/.github/workflows/inference_cache.yml
@@ -0,0 +1,61 @@
+name: Optimum neuron inference cache builder
+
+on:
+  workflow_dispatch:
+  schedule:
+    # Schedule the workflow to run every day at midnight UTC
+    - cron: '0 0 * * *'
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
+
+jobs:
+  cache:
+    name: Create optimum-neuron inference cache
+    runs-on: [self-hosted, 12-aws-inf2, 192-cpu, ci] # run the job on the newly created runner
+    env:
+      AWS_REGION: us-east-1
+    strategy:
+      fail-fast: false
+      matrix:
+        config: [
+          gpt2,
+          llama3-8b,
+          llama3-70b,
+          llama2-7b-13b,
+          llama2-70b,
+          mistral,
+          llama-variants,
+          mistral-variants,
+          mixtral
+        ]
+    steps:
+      - name: Install Neuron runtime
+        run: |
+          . /etc/os-release
+          sudo tee /etc/apt/sources.list.d/neuron.list > /dev/null <<EOF
+          deb https://apt.repos.neuron.amazonaws.com ${VERSION_CODENAME} main
+          EOF
+          wget -qO - https://apt.repos.neuron.amazonaws.com/GPG-PUB-KEY-AMAZON-AWS-NEURON.PUB | sudo apt-key add -
+          sudo apt-get update -y
+          sudo apt-get install aws-neuronx-tools=2.17.1.0 aws-neuronx-runtime-lib=2.20.22.0-1b3ca6425 aws-neuronx-collectives=2.20.22.0-c101c322e  -y
+          export PATH=/opt/aws/neuron/bin:$PATH
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Install python and create venv
+        run: |
+          sudo apt install python3.8-venv python3-dev -y
+          python3 -m venv aws_neuron_venv_pytorch
+          source aws_neuron_venv_pytorch/bin/activate
+          python -m pip install -U pip
+          python -m pip config set global.extra-index-url https://pip.repos.neuron.amazonaws.com
+      - name: Install optimum neuron
+        run: |
+          source aws_neuron_venv_pytorch/bin/activate
+          python -m pip install .[neuronx]
+      - name: Create cache for ${{matrix.config}} models
+        run: |
+          source aws_neuron_venv_pytorch/bin/activate
+          config_prefix_url=https://huggingface.co/aws-neuron/optimum-neuron-cache/raw/main/inference-cache-config
+          HF_TOKEN=${{secrets.HF_TOKEN_OPTIMUM_NEURON_CACHE}} \
+            python tools/auto_fill_inference_cache.py --config_file ${config_prefix_url}/${{matrix.config}}.json