inference/docker/publish/README.md

0. Get added to roboflow dockerhub
1. Bump verison in inference/core/version.py
2. docker login
3. cd to root
4. To deploy x86 images: ./deploy/cpu_http.sh ./deploy/gpu_http.sh ./deploy/trt_http.sh
5. To deploy aws Lambda images to [staging|platform]: source ~/aws_keys/[staging|platform].sh
   • first time logging into staging ECR:
     • aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin 809383754475.dkr.ecr.us-east-1.amazonaws.com
   • first time logging into prod ECR:
     • aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin 583392192139.dkr.ecr.us-east-1.amazonaws.com
   • Upload container: ./deploy/onnx_lambda.sh roboflow-[staging|platform]
   • Go to roboflow-infra and change Lambda tags
   • NOTE: Test out staging deploy before deploying prod
6. To deploy ARM images:
   • Go to ARM VM: gcloud compute ssh --zone "us-central1-a" "arm-cpu-dev-machine" --project "roboflow-staging"
   • docker login
   • pull code: git clone git@github.com:roboflow/inference.git
   • ./deploy/arm_cpu.http ./deploy/jetson_trt_http.sh ./deploy/jetson_trt_http_5.1.1.sh