Stacked @card v1.2 : New MetaflowCardComponents

[valayDave]

Stacked on : #895

What are the Changes :

• introducing user facing MetaflowCardComponents
• import via from metaflow.cards import Artifact,..
• exposing public methods in TaskToDict
• remove imports from metaflow.cards
• Tests for the Card UI

Example Flow to Test these:

This flow requires pandas ,numpy and PIL installed. You can use the components without those too.

from metaflow import FlowSpec,step,batch,current,Parameter,card
import requests
class ModelTrainingFlow(FlowSpec):
   
    @step
    def start(self):
        self.next(self.train)
        

    @card
    @step
    def train(self):
        import random
        import numpy as np
        self.exec_medium = "local"
        self.train_model()
        self.loss = np.random.randn(100,100)*100
        self.next(self.end)
    
    def train_model(self):
        import pandas as pd
        import numpy as np
        from metaflow.cards import Artifact,Image,Table,Section,Error
        from metaflow import current
        from PIL import Image as PILImage
        df = pd.DataFrame(np.random.randint(0,100,size=(15, 4)), columns=list('ABCD'))
        plot = df.plot()
        components = [
                Artifact(df,'Dataframe as Artifact',),
                Image.from_matplotlib(plot,"Image From user code",'my dummy plot'),
                Table.from_dataframe(df,"my core df"),
                Image(requests.get('https://www.gif-vif.com/hacker-cat.gif').content,"Cat image"),
                Error(Exception('something went wrong'),'Something excepted')
            ]
        current.card.extend(
            components
        )
        current.card.append(
            Section(components,"Main Section under which everything comes")   
        )
    
    @step
    def end(self):
        print("Done Computation")

if __name__ == "__main__":
    ModelTrainingFlow()