PyTorch Serverless

FastAI PyTorch Serverless API (w/ AWS Lambda)

Setup

• Install Serverless Framework via npm

  npm i -g serverless@v1.27.3

• Install python requirements plugin

  sls plugin install -n serverless-python-requirements

Configuration

• Setup your model in lib/models.py so that it can be imported by the handler in api/predict.py as a method

• Define your class labels in lib/labels.txt with one label per line, for example:

  cat
  dog

• Setup an AWS CLI profile if you don't have one already

• Create an S3 Bucket that your profile can access and upload your state dictionary

• Configure the serverless.yml

  ### Change service name to whatever you please
  service: pytorch-serverless
  
  provider:
      ...
      ### set this to your deployment stage
      stage: dev
  
      ### set this to your aws region
      region: us-west-2
  
      ### set this to your aws profile
      profile: slsadmin
  
      ### set this as needed between 128 - 3008, in 64mb intervals
      memorySize: 2048
  
      ### set this as needed (max 300)
      timeout: 120
      ...
  
      environment:
          ### set this to your S3 bucket name
          BUCKET_NAME: pytorch-serverless
  
          ### set this to your state dict filename
          STATE_DICT_NAME: dogscats-resnext50.h5
  
          ### set this to your input image size
          IMAGE_SIZE: 224
  
          ### set this to your image normalization stats
          IMAGE_STATS: ([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
          ...
  
      variables:
          ### set this to your api version
          api_version: v0.0.1

Invoke Local

Run function locally with params defined in tests/predict_event.json

AWS_PROFILE=yourProfile sls invoke local -f predict -p tests/predict_event.json

Deployment

Make sure Docker is running

Deploy to AWS Lambda

sls deploy -v

Endpoints

GET /predict

Return prediction for a single image.

• Headers

  (required)
  X-API-KEY=[string]   ### Your generated API Key

• URL Parameters

  (required)
  image_url=[url]   ### URL of image to classify
  
  (optional)
  top_k=[integer]   ### Number of top results (default: 3)

• Success Response (200)

  {
      "predictions": [
          {
  
            "label": "dog",
            "log": -0.00004426980376592837,
            "prob": 0.9999557137489319
          },
          {
  
            "label": "cat",
            "log": -10.025229454040527,
            "prob": 0.0000442688433395233
          }
      ]
  }

• Error Response (500)

  {
      "error": "Something went wrong...",
      "traceback": "..."
  }

Logs

Tail logs to console

sls logs -f predict -t