Incremental training for Deep Learning and Wapiti models #971

kermitt2 · 2022-11-24T14:19:03Z

Support incremental training for Deep Learning models.

We simply add -i to the training command, and the training will resume from the existing model instead of starting from scratch with a new model.

> java -Xmx1024m -jar grobid-trainer/build/libs/grobid-trainer-<current version>-onejar.jar 0 <name of the model> -gH grobid-home -i

It does work for CRF Wapiti too.

coveralls · 2022-11-24T16:37:33Z

Coverage decreased (-0.08%) to 39.586% when pulling fc701f7 on incremental-training into b9d92c6 on master.

kermitt2 added 3 commits November 24, 2022 11:38

support incremental training for DL models

2dce9b9

document incremental training call

b1e4fd1

add incremental training for Wapiti

3209bbb

kermitt2 changed the title ~~Incremental training for DL models~~ Incremental training for Deep Learning and Wapiti models Nov 24, 2022

kermitt2 added 4 commits November 25, 2022 09:09

some missing steps in JNI training (although not used)

b61eb87

ensure compatibility with older modules/versions

b166763

add option incremental in REST web training service

c6b33a7

document the incremental training option in the web api training service

fc701f7

kermitt2 merged commit 2bfadfd into master Nov 25, 2022

Provide feedback