search

lancercat / VSDF

23 stars 5 forks source link

How to specify multi - gpu training? #3

Closed 1193700079 closed 2 years ago

1193700079 commented 2 years ago

The default setting is to start only one gpu

1193700079 commented 2 years ago

by the way, image

these variables: choutput and ctxout tarswunk and gtarswunk

What does that mean？

lancercat commented 2 years ago

  1. Yup, it's designed to use only one GPU. MultiGPU will NOT be implemented in this work.

  2. variables

    choutput: prediction from the vision module

    ctxout: prediction from the semantic module

    tarswunk: local labels(TARgetS) in batch With UNKnown labels replaced(as the label sampler samples label)

    gtarswunk: Global labels(TARgetS) With UNKnown labels replaced for semantic module.

1193700079 commented 2 years ago

ok thank you !