Thomas Dybdahl Ahle

Something like cutechess' 40/40+1 notation could be nice.

Allow parallel function calls

2

Is there a way to run the objective function in parallel? Ideally something like the "ask and tell" interface to skopt: https://scikit-optimize.github.io/auto_examples/ask-and-tell.html

Documentation for `demo_train.py`

1

I'm trying to run `demo_train.py`, but it's unclear how it works. If I add the `--glove` parameter, it complains that both sift and glove is active, since sift is apparently...

Json Support

9

Is it possible to use json objects for context and answers? Is the best solution to do something like: ```python answer = dspy.OutputField(desc=""" Use the following format: {"summary": string, "commands:...

Docstrings

2

Having doc-strings on methods and classes would be very helpful. E.g., I'm currently trying to figure out how the `parse_structure`, `attach` and `add_field` methods on `Signature` work. But they are...

Evaluate should take multiple metrics

3

Right now `Evaluate(...)` only take one metric, but often we have multiple different scores we want to test at the same time. Like "accuracy" and "gold_passages_retrieved" and "q/s" etc. While...

I'm trying to use the following prompt: ``` next_steps = dspy.OutputField( desc=""" List of 2-5 promising next steps. Use the format: x op y = z (left: ns). - x...

New Signature class based on pydantic + unit tests

4

This PR creates a new Signature class based on Pydantic, which allows for type signatures. It doesn't actually do anything with the types for now, but it's the first step....

dspy.Predict should be a dspy.Module

9

For most purposes `dspy.Predict` behaves the same was as a `dspy.Module`. But if you try to pass a `Predict` directly to an optimizer, you'll notice that it's lacking a lot...

Define default formatting

1

Using an auto-formatting is a great way to avoid stylistic conflicts/flame-wars in a repository. It also makes it easier for new comers, because the format is similar to what they've...