Skip to content
This repository has been archived by the owner on Feb 3, 2023. It is now read-only.

Train MLP

Jump to bottom
James Thorne edited this page Jul 5, 2018 · 1 revision

Model 1: Multilayer Perceptron

#If using a GPU, set
export GPU=1
#If more than one GPU,
export CUDA_DEVICE=0 #(or any CUDA device id. default is 0)

# Using nearest neighbor sampling method for NotEnoughInfo class (better)
PYTHONPATH=src python src/scripts/rte/mlp/train_mlp.py data/fever/fever.db data/fever/train.ns.pages.p1.jsonl data/fever/dev.ns.pages.p1.jsonl --model ns_nn_sent --sentence true

#Or, using random sampled data for NotEnoughInfo (worse)
PYTHONPATH=src python src/scripts/rte/mlp/train_mlp.py data/fever/fever.db data/fever/train.ns.rand.jsonl data/fever/dev.ns.rand.jsonl --model ns_rand_sent --sentence true
Clone this wiki locally