todo.md

• model class needs to be able to run a given pytorch model
• instantly train 1 iteration given set of data points
• dataset to act as transformer

# Method that retruns dataloader, model
builder(train_info):
    for datapoint in train_info["data"]:
        datapoint["data"]
        datapoint["label"]

    => create a dataset => dataloader

    model(train_info["model_conf"])
    return dataloader, model

-------------
# Generic that returns loss and model from training
trainer(dataloader, model):
    model.optimizer
    training loop here
    return cirterion, model

-------------

main(train_info):
    dataloader, model = builder(train_info)
    cirterion, model = trainer(dataloader, model)
    model.save()
    return cirterion

Trainer

• Desc Object to handle download of data, model/dataset creatation and training of model
• Method
  • pull_data consider it in init?
  • create dataloader
  • create model

Model

• Instance Vars

• Methods

  • Train method