GitXplorerGitXplorer
w

PixieVGModel

public
2 stars
0 forks
0 issues

Commits

List of commits on branch main.

No commits found

There are no commits on branch main.

README

The README file for this repository.

I'm still cleaning up this repo. Will release the final version soon.

PixieModel

Evaluation datasets

MEN: https://aclweb.org/aclwiki/MEN_Test_Collection_(State_of_the_art)

SimLek-999: https://fh295.github.io/simlex.html

Generate dataset from Visual Genome

--min_freq to set the minimum threshold of triple occurrence.

--pixie_dim to set the dimension of pixie, which is 100.

--data_pach

--pca-path

python3 preprocessing.py \
    --VG_path=/the_path_of_your_VG_zip_data/ \
    --pixie_dim=100 \
    --data_path=/the_path_of_your_filtered_transformed_data/ \
    --pca_path=/the_path_of_your_pca_transformed_data/ \
    --pca_only=True \
    --min_freq=100

Train the PixieModel

--pixie_dim

--data_path

--parameter_path

--pca_path

--lr

--dr

epoch_num

python3 pixie_model.py \
    --pixie_dim=100 \
    --pca_path=/the_path_of_your_pca_transformed_data/ \
    --parameter_path='parameters/' \
    --lr=0.01 \
    --dr=5e-8 \
    --epoch_num=20

Evalute the PixieModel

--pixie_dim: Dimension of the pixie, default is 100.

--dataset: This could be one of the 'MEN', 'Simlek', 'RELPRON' and 'GS2011'.

--data_path

--pca_path

--parameter_path: The path of the trained parameters 'world_parameters.p' and 'Lexical_parameters.p'.

--use_EVA_vocab: Boolean value to set if only use the vocab covered by the EVA work.

python3 evaluation.py --dataset=Simlek --pixie_dim=100 --data_path=pixie_data_2/ --pca_path=data_pca_2/