Thomas Capelle

Update test.yaml

1

Ignore tests if the changes are only in the docs

[WIP] Anthropic integration

2

- Add basic weave.op support for anthropic SDK

[WIP] Simplify and refactor to use Weave

3

- A simpler finetune script - An eval script that uses weave and Evaluations - Better prompt defaults

feat: integration Mistral V1.0 support

2

This adds support for the newly released Mistral v1.0 SDK: https://github.com/mistralai/client-python/blob/main/MIGRATION.md - Moves mistralai

feat(integration): Add Anthropic Messages.stream support

4

Antrhopic has 2 types of streaming: One by event and another by messages. We are missing the `Messages.stream` support, this PR tries to add this. On this mode, the antrhopic...

docs(best-practices) Some recipes that are useful

4

- Instrumenting with weave - Weave and Pydantic: Model, Object, etc... - Using `model_post_init` and `PrivateAttr`, for delayed attributes - Example of using this for `client` and a `HF` model.

feat(Improve Datasets)

1

The dataset class is still very lightweight but has a lot of potential. More now that we will have feedback and ways to annotate data. Let's try to put some...

Simple HF model wrapping

2

Adds a doc page on how to wrap a local HF model with Weave

feat(weave): Adds Dataset.map and progress bars

4

This PR replaces #3751 - Modifies `async_foreach` to support rich progressbar - Enables `weave.Evaluation` with progress bar, also `verbose=False` to silence printing per row - Adds `Dataset.map` that uses `async_foreach`...

feat(weave): HF.Datasets <-> weave.Dataset

1

Adds simple `to_hf` and `from_hf` methods to `weave.Dataset` so we can do the processing leveraging the fast and battle tested datasets library.