Lesson Video:

Starlette
Follow the Render tutorial here
Link to a fastai template
Note: You do not need to deploy on Render to get the code working, we can test locally on our machine! (which we will do today)

What will we focus on?

Looking at how to format inputs/outputs for each model type and feeding it in.
Images, Tabular, NLP

What code do we change?

server.py

Images:

Different input types:
- URL
- File upload

async def get_bytes(url):
  async with aiohttp.ClientSession() as session:
    async with session.get(url) as response:
      return await response.read()

An image upload

@app.route('/analyze', methods=['POST'])
async def analyze(request):
  img_data = await request.form()
  img_bytes = await (img_data['file'].read())
  pred = learn.predict(img_bytes)[0]
  return JSONResponse({
      'results': str(pred)
  })

A URL

@app.route('/analyze', methods=['POST'])
async def analyze(request):
  img_bytes = await get_bytes(request.query_params["url"])
  pred = learn.predict(img_bytes)[0]
  return JSONResponse({
      'results' : str(pred)
  })

A zip file (see below on how to upload a zip or other file

import zipfile
import csv

@app.route('/analyze', methods=['POST'])
async def analyze(request):
  data = await request.form()
  content = data['content']
  zip_ref = zipfile.ZipFile(content, 'r')
  mkdir('Downloaded_Images')
  zipref.extractall('Downloaded_Images')
  zip_ref.close()
  path = Path('Downloaded_Images')
  imgs = get_image_files(path)
  learn = load_learner(path/export_file_name)
  dl = test_dl(learn.dls, imgs)
  _, __, preds = learn.get_preds(dl=dl, with_decoded=True)
  rm -r 'Downloaded_Images'
  resultsFile = open('results.csv', 'wb')
  wr = csv.writer(resultsFile)
  wr.writerows([preds])
  return FileResponse('results.csv')

Parsing a csv with image urls

import csv
import StringIO

@app.route('/analyze', methods=['POST'])
async def analyze(request):
  data = await request.form()
  content = await (data['file'].read())
  s = str(content, 'utf-8')
  data = StringIO(s)
  mkdir('Downloaded_Images')
  download_images('Downloaded_Images', urls=data)
  path = Path('Downloaded_Images')
  learn = load_learner(path/export_file_name)
  imgs = get_image_files(path)
  dl = test_dl(learn.dls, imgs)
  _, __, preds = learn.get_preds(dl=dl, with_decoded=True)
  rm -r 'Downloaded_Images'
  resultsFile = open('results.csv', 'wb')
  wr = csv.writer(resultsFile)
  wr.writerows([preds])
  return FileResponse('results.csv')

Tabular

Tabular is different. Most work will be done by sending large chuncks of data for analysis. Let's recreate what we did, but load it into Pandas

import StringIO
import csv

@app.route('/analyze', methods=['POST'])
async def analyze(request):
  data = await request.form()
  content = await (data['file'].read())
  s = str(content, 'utf-8')
  data = StringIO(s)
  df = pd.read_csv(data)
  learn = load_learner(path/export_file_name)
  # if we want to do GPU:
  # learn.model = learn.model.cuda()
  dl = learn.dls.train_dl.new(df)
  _, __, y = learn.get_preds(dl=dl, with_decoded=True)
  df['Predictions'] = y
  # if we want to store the results
  path_res = Path('app/static/')
  df.to_csv(path_res/'results.csv')

  return FileResponse('results.csv', media_type='csv')

We need to adjust the JavaScript to accept a form:

client.js:

function analyze(){
    var uploadFiles = el('file-input').files;
    if (uploadFiles.length < 1) alert('Please select 1 file to analyze!');

    el('analyze-button').innerHTML = 'Analyzing...';
    var xhr = new XMLHttpRequest();
    var loc = window.location
    xhr.open('POST', `${loc.protocol}//${loc.hostname}:${loc.port}/analyze`, true);
    xhr.onerror = function() {alert (xhr.responseText);}
    xhr.onload = function(e) {
        if (this.readyState === 4) {
          el("result-label").innerHTML = `Result = Good`;
          
          download('results.csv', 'results.csv');
          xhr.send();
        }
        el("analyze-button").innerHTML = "Analyze";
      };

      var fileData = new FormData();
      fileData.append("file", uploadFiles[0]);
      xhr.send(fileData);
    }
  }

Text

To write a simple function for text based models: