Skip to content

Run Inference in Parallel #114

Run Inference in Parallel

Run Inference in Parallel #114

name: Run Inference in Parallel
on:
workflow_dispatch:
inputs:
model-id:
required: true
type: string
sample-only:
required: false
type: string
n-workers:
description: 'number of workers to use (max 50)'
required: true
default: '50'
type: string
permissions:
contents: read
jobs:
start-time:
runs-on: ubuntu-latest
outputs:
start-time: ${{ steps.start-time.outputs.start-time }}
steps:
- name: Get start time
id: start-time
run: echo "start-time=$(date +%s)" >> "$GITHUB_OUTPUT"
pull-model-image:
runs-on: ubuntu-latest
steps:
- name: Remove unnecessary files
run: |
sudo rm -rf /usr/share/dotnet
sudo rm -rf "$AGENT_TOOLSDIRECTORY"
sudo rm -rf /opt/ghc
sudo rm -rf "/usr/local/share/boost"
- name: cache model Docker image
id: cache
uses: actions/cache@v3
with:
path: /tmp/${{ inputs.model-id }}.tar
key: ${{ runner.os }}-docker-${{ inputs.model-id }}
restore-keys: |
${{ runner.os }}-docker-${{ inputs.model-id }}
- name: pull and save requested model image
if: steps.cache.outputs.cache-hit != 'true'
run: |
docker pull ersiliaos/${{ inputs.model-id }}:latest
docker save ersiliaos/${{ inputs.model-id }} -o /tmp/${{ inputs.model-id }}.tar
generate-matrix:
needs: pull-model-image
runs-on: ubuntu-latest
outputs:
matrix: ${{ steps.set-matrix.outputs.matrix }}
steps:
- id: set-matrix
run: |
start='1'
end=${{ inputs.n-workers }}
matrix=$(seq -s ',' $start $end)
echo "matrix=[${matrix}]" >> $GITHUB_OUTPUT
matrix-inference:
needs: generate-matrix
if: github.repository != 'ersilia-os/eos-template'
strategy:
matrix:
numerator: ${{ fromJson(needs.generate-matrix.outputs.matrix) }}
uses: ./.github/workflows/predict.yml
with:
numerator: ${{ matrix.numerator }}
denominator: ${{ inputs.n-workers }}
model-id: ${{ inputs.model-id }}
sample-only: ${{ inputs.sample-only }}
SHA: ${{ github.sha }}
secrets: inherit