Commit graph

13 commits

Author SHA1 Message Date
Xi Yan
a4bcfb8bba
[/scoring] add ability to define aggregation functions for scoring functions & refactors (#597)
# What does this PR do?

- Add ability to define aggregation functions for scoring functions via
`ScoringFnParams`
- Supported by `basic` / `regex_parser` / `llm_as_judge` scoring
functions


## Test Plan

```
pytest -v -s -m basic_scoring_together_inference scoring/test_scoring.py
```
<img width="855" alt="image"
src="https://github.com/user-attachments/assets/12db8e6e-2ad4-462e-b9b9-70ba6c050a6c">


```
pytest -v -s -m llm_as_judge_scoring_together_inference scoring/test_scoring.py
```
<img width="858" alt="image"
src="https://github.com/user-attachments/assets/bf806676-6f5e-456d-be9f-f81a26d1df19">



**Example Response** (`basic`)
<img width="863" alt="image"
src="https://github.com/user-attachments/assets/0e57a49c-8386-45cc-8fa9-3e61aaa9a3be">

**Example Response** (`llm-as-judge`)
<img width="854" alt="image"
src="https://github.com/user-attachments/assets/38065bc2-b724-47ed-9535-79b6099c4362">


## Sources

Please link relevant resources if necessary.


## Before submitting

- [ ] This PR fixes a typo or improves the docs (you can dismiss the
other checks if that's the case).
- [ ] Ran pre-commit to handle lint / formatting issues.
- [ ] Read the [contributor
guideline](https://github.com/meta-llama/llama-stack/blob/main/CONTRIBUTING.md),
      Pull Request section?
- [ ] Updated relevant documentation.
- [ ] Wrote necessary unit or integration tests.
2024-12-11 10:03:42 -08:00
Ashwin Bharambe
0dc7f5fa89
Add version to REST API url (#478)
# What does this PR do? 

Adds a `/alpha/` prefix to all the REST API urls.

Also makes them all use hyphens instead of underscores as is more
standard practice.

(This is based on feedback from our partners.)

## Test Plan 

The Stack itself does not need updating. However, client SDKs and
documentation will need to be updated.
2024-11-18 22:44:14 -08:00
Ashwin Bharambe
983d6ce2df
Remove the "ShieldType" concept (#430)
# What does this PR do?

This PR kills the notion of "ShieldType". The impetus for this is the
realization:

> Why is keyword llama-guard appearing so many times everywhere,
sometimes with hyphens, sometimes with underscores?

Now that we have a notion of "provider specific resource identifiers"
and "user specific aliases" for those and the fact that this works with
models ("Llama3.1-8B-Instruct" <> "fireworks/llama-3pv1-..."), we can
follow the same rules for Shields.

So each Safety provider can make up a notion of identifiers it has
registered. This already happens with Bedrock correctly. We just
generalize it for Llama Guard, Prompt Guard, etc.

For Llama Guard, we further simplify by just adopting the underlying
model name itself as the identifier! No confusion necessary.

While doing this, I noticed a bug in our DistributionRegistry where we
weren't scoping identifiers by type. Fixed.

## Feature/Issue validation/testing/test plan

Ran (inference, safety, memory, agents) tests with ollama and fireworks
providers.
2024-11-12 12:37:24 -08:00
Ashwin Bharambe
09269e2a44
Enable sane naming of registered objects with defaults (#429)
# What does this PR do? 

This is a follow-up to #425. That PR allows for specifying models in the
registry, but each entry needs to look like:

```yaml
- identifier: ...
  provider_id: ...
  provider_resource_identifier: ...
```

This is headache-inducing.

The current PR makes this situation better by adopting the shape of our
APIs. Namely, we need the user to only specify `model-id`. The rest
should be optional and figured out by the Stack. You can always override
it.

Here's what example `ollama` "full stack" registry looks like (we still
need to kill or simplify shield_type crap):
```yaml
models:
- model_id: Llama3.2-3B-Instruct
- model_id: Llama-Guard-3-1B
shields:
- shield_id: llama_guard
  shield_type: llama_guard
```

## Test Plan

See test plan for #425. Re-ran it.
2024-11-12 11:18:05 -08:00
Dinesh Yeduguru
0a3b3d5fb6
migrate scoring fns to resource (#422)
* fix after rebase

* remove print

---------

Co-authored-by: Dinesh Yeduguru <dineshyv@fb.com>
2024-11-11 17:28:48 -08:00
Xi Yan
6192bf43a4
[Evals API][10/n] API updates for EvalTaskDef + new test migration (#379)
* wip

* scoring fn api

* eval api

* eval task

* evaluate api update

* pre commit

* unwrap context -> config

* config field doc

* typo

* naming fix

* separate benchmark / app eval

* api name

* rename

* wip tests

* wip

* datasetio test

* delete unused

* fixture

* scoring resolve

* fix scoring register

* scoring test pass

* score batch

* scoring fix

* fix eval

* test eval works

* remove type ignore

* api refactor

* add default task_eval_id for routing

* add eval_id for jobs

* remove type ignore

* only keep 1 run_eval

* fix optional

* register task required

* register task required

* delete old tests

* delete old tests

* fixture return impl
2024-11-07 21:24:12 -08:00
Dinesh Yeduguru
663883cc29
persist registered objects with distribution (#354)
* persist registered objects with distribution

* linter fixes

* comment

* use annotate and field discriminator

* workign tests

* donot use global state

* precommit failures fixed

* add back Any

* fix imports

* remove unnecessary changes in ollama

* precommit failures fixed

* make kvstore configurable for dist and rename registry

* add comment about registry list return

* fix linter errors

* use registry to hydrate

* remove debug print

* linter fixes

* remove kvstore.db

* rename distribution_registry_store

---------

Co-authored-by: Dinesh Yeduguru <dineshyv@fb.com>
2024-11-04 17:25:06 -08:00
Ashwin Bharambe
26d1668f7d Revert "remove Field for return_type"
This reverts commit ffb3965ade.
2024-10-28 21:39:48 -07:00
Xi Yan
7b8748c53e
[Evals API][6/n] meta-reference llm as judge, registration for ScoringFnDefs (#330)
* wip scoring refactor

* llm as judge, move folders

* test full generation + eval

* extract score regex to llm context

* remove prints, cleanup braintrust in this branch

* change json -> class

* remove initialize

* address nits

* check identifier prefix

* udpate MANIFEST
2024-10-28 14:08:42 -07:00
Xi Yan
ffb3965ade remove Field for return_type 2024-10-28 13:04:41 -07:00
Xi Yan
abdf7cddf3
[Evals API][4/n] evals with generation meta-reference impl (#303)
* wip

* dataset validation

* test_scoring

* cleanup

* clean up test

* comments

* error checking

* dataset client

* test client:

* datasetio client

* clean up

* basic scoring function works

* scorer wip

* equality scorer

* score batch impl

* score batch

* update scoring test

* refactor

* validate scorer input

* address comments

* evals with generation

* add all rows scores to ScoringResult

* minor typing

* bugfix

* scoring function def rename

* rebase name

* refactor

* address comments

* Update iOS inference instructions for new quantization

* Small updates to quantization config

* Fix score threshold in faiss

* Bump version to 0.0.45

* Handle both ipv6 and ipv4 interfaces together

* update manifest for build templates

* Update getting_started.md

* chatcompletion & completion input type validation

* inclusion->subsetof

* error checking

* scoring_function -> scoring_fn rename, scorer -> scoring_fn rename

* address comments

* [Evals API][5/n] fixes to generate openapi spec (#323)

* generate openapi

* typing comment, dataset -> dataset_id

* remove custom type

* sample eval run.yaml

---------

Co-authored-by: Dalton Flanagan <6599399+dltn@users.noreply.github.com>
Co-authored-by: Ashwin Bharambe <ashwin.bharambe@gmail.com>
2024-10-25 13:12:39 -07:00
Xi Yan
cb84034567
[Evals API][3/n] scoring_functions / scoring meta-reference implementations (#296)
* wip

* dataset validation

* test_scoring

* cleanup

* clean up test

* comments

* error checking

* dataset client

* test client:

* datasetio client

* clean up

* basic scoring function works

* scorer wip

* equality scorer

* score batch impl

* score batch

* update scoring test

* refactor

* validate scorer input

* address comments

* add all rows scores to ScoringResult

* bugfix

* scoring function def rename
2024-10-24 14:52:30 -07:00
Xi Yan
e45f121c77
[Evals API] [1/n] Initial API (#287)
* type system api

* datasets api

* fix

* datasetio api

* kill reward scoring

* scoring functions + evals

* move jobs, fix errors
2024-10-22 09:31:19 -07:00