OpenNMT / OpenNMT/CTranslate2

faster_whisper batch operation time-consuming

Open
#1,588 9 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
C++
Stars
4.7k
Forks
536
Avg merge
12h 12m
Merged PRs (30d)
4

Description

Is this related to CTranslate?

The following is copied from this .
I have made a test, for batching in faster-whisper.

But faster_whisper batch encode consume multiple time as sample's amount, it seems encode in batch not work as expected by CTranslate?

By the way, the decode/generate have the same issue when in-Batch ops.
290427349-927b5fdc-d58b-45a2-87f3-0da80d3a5ac3

290427947-e0728401-4d04-4373-85c0-23bb59c1d7af

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reproducing the reported faster-whisper batch encode and decode/generate timings against the CTranslate2 behavior described in the issue and the linked faster-whisper report. Compare batched operations with equivalent single-sample operations and determine whether the observed scaling is expected; done means the behavior is explained or a focused fix is validated with timing evidence.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp
Domain
machine-learning, performance
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.