OpenNMT / OpenNMT/CTranslate2

Get encoding from flan T5

Open
#1,337 1 comment 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

Hi, I want to use this lib to get encodings (on all positions) from flan T5 encoder on CPU. But I am not familiar with c++, so it is hard to do the customization myself. Is it possible to support it?

Just one API on Translator to return batch size * length * hidden_dimension encodings works. Thanks!

Context:
I am trying to train text-to-image model, and I want to put text encoder on CPU. I think this implementation is the best option I can find to support T5 encoder on CPU, with speed optimization.

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

The request concerns the Translator API and the Flan-T5 encoder, with encodings returned for every position on CPU. Start by locating the Translator entry point and the encoder implementation; done means a documented API returns batch-size × length × hidden-dimension encodings for Flan-T5 on CPU.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp
Domain
machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.