`Dataset` in `etna.datasets`

[martins0n]

🚀 Feature Request

class Dataset:
    train: TSDataset
    test: TSDataset
    dataset_path: str # url or repositary url 
    freq: str
    known_future: Optional[List[Feature]]
    unknown_future: Optional[List[Feature]]
    cache_path: Optional[pathlib.Path]
    metadata: dict
    tags: List[str]
  
   @property
   def train(self) -> TSDataset:
        pass

   @property
   def test(self) -> TSDataset:
        pass

class Feature:
   # N.B. its just possible fields -- we shoud use dict instead of classes
   name: str
   type: Union[Literal["categorical"], Literal["numeric"], Literal["str"]]

def m5_generation() -> Dataset:
   # code for preprocessing of local saved dataset

def load_dataset(name: str) -> Dataset:
   # calling function for generation Dataset

• [ ] Index folder with structure:

etna/datasets/index/m4_monthly.json
...
etna/datasets/index/m5.json

With json configuration for TSDataset creation it could contains specials params for data generation, urls, fields for Dataset init

• [ ] We have helper function which produce Dataset using information in json.

• [ ] caching datasets in jsonl #467 format (Optional, in the first itreation we can download dataset everytime)

• [ ] we should add current code generated datasets and m5

Motivation

Proposal

Test cases

No response

Alternatives

No response

Additional context

#467

Checklist

• [ ] I discussed this issue with ETNA Team