TODO from discussion

[rstojnic]

TODO sotabench lib:

• remove benchmark() function from benchmark.py
• move deps to requirements
• evaluation.json should be made if some ENV variable is set, otherwise pprint something
• for each benchmark:
  • benchmark()
  • default transform
  • the dataset
  • default parameters
• documentation:
  • dataset examples
  • default transform example
  • input fed to model, and expected output
  • link to examples of benchmarked models
• a library of transforms (maybe)

And additional requests:

• BenchmarkResult return value should also contain: 1) the dataset used, 2) the transform used, 3) input parameters used when invoking the function, 4) anything else - so it's a self-contained record of results