Skip to content

parafoxia/decks

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

DECKS

The Dimensional Emotion-Contextualising Kinetic System.

Prerequisits

  • docker

Training DECKS

Clone this repo, then build the Docker image:

docker build -t decks:<gen> .

To train models:

docker run --gpus <devices> --ipc=host -it decks:<gen> python -m decks <gen> <-e epochs> [-b batch_size] [-i epoch]

Arguments:

  • devices - A Docker flag for GPU devices. See below for more information.
  • gen - The model generation.
  • -e epochs - The number of epochs to train for.
  • -b batch_size (optional) - The size of data batches.
  • -i epoch (optional) - Continue training from the given epoch. The generation, epochs, and batch size must have been used for a previous model, otherwise an error will be raised.

Note: ipc=host can be omitted if only one GPU device is made available.

Specifying devices

Use case Command
Only use GPU 0 --gpus '"device=0"'
Use GPUs 1 and 3 --gpus '"device=1,3"'
Use all GPUs --gpus all

Note: If more than one GPU device is made available, distributed training is automatically enabled.

Evaluating DECKS

Several notebooks have been provided in the notebooks directory. These will need to moved to the root directory to run.

About

Dimensional Emotion-Contextualising Kinetic System

Resources

Stars

Watchers

Forks