David Nicholson

I got this to a point where it's working. There's an issue, though. We do the post-processing and then return the labels all in a single call to a single...

picking this up again what we want: - a way to get a `Model` instance for any model we declare where we can do two things: - just get the...

Here's a code snippet that (I think) illustrates what I have in mind: ```python def model(model): """a decorator that creates a model""" class Model: self._model = model def __init__(network, loss,...

Unfinished business: - I want to be able to do the following I get a new model with defaults, so I don't need to pass in any config ```python VakMyModel...

I think I have a very Minimal VP of this working: ```python import functools import vak import torch import tweetynet def model(model): """a decorator that creates a model""" @functools.wraps(model, updated=())...

Feature branch is already in progress, but documenting here how I ended up implementing this. Basically, as follows: - [ ] Have an `attrs`-like or `dataclasses`-like `ModelDefinition` that specifies a...

Here's the Minimum Viable Implementation of that: ```python import functools import pathlib from typing import Callable, ClassVar, NewType import lightning import torch import vak from vak import labeled_timebins class ModelDefinition:...

Closed by #605

Hi @vivinastase I added some initial info on how to run eval just now in #715 (a year and a half later :sob:) -- leaving this open because I know...

@all-contributors please add @vivinastase for ideas