huggingface / huggingface/evaluate

Allow BLEURT optimisation parameters

Open
#515 0 comments 1 reaction 0 assignees View on GitHub
Dominant language
Python
Stars
2.5k
Forks
341
PR merge metrics
No merged PRs in 30d

Description

BLEURT is very slow to compute: I'm observing 426.62 seconds on a g5.xlarge with `BLEURT-20-D6` for 1459 sentences.

It'd be useful to be able to use the optimisations specified in [speeding-up-bleurt](https://github.com/google-research/bleurt/tree/master#speeding-up-bleurt)

1. Tune batch size by passing `batch_size` as a kwarg to `scorer.score` (https://github.com/google-research/bleurt/blob/cebe7e6f996b40910cfaa520a63db47807e3bf5c/bleurt/score.py#L175)
2. Use the `LengthBatchingBleurtScorer` (https://github.com/google-research/bleurt/blob/cebe7e6f996b40910cfaa520a63db47807e3bf5c/bleurt/score.py#L227C7-L227C33)

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.