- Install PyTorch however you want
- Install requirements
pip install -r requirements.txt
- Download doc2vec here and place files in
data/doc2vec/
- Download the dataset here and place files in
data/memes/
- Train MemeNet by running
python -m memenet.train
in the same directory asREADME.md
- Run streamlit app
streamlit run app.py
If something is not working, first ensure that the data/
directory looks like this:
data
├── doc2vec
│ ├── doc2vec_wiki_d300_n5_w8_mc50_t12_e10_dbow.model
│ ├── doc2vec_wiki_d300_n5_w8_mc50_t12_e10_dbow.model.docvecs.vectors_docs.npy
│ ├── doc2vec_wiki_d300_n5_w8_mc50_t12_e10_dbow.model.trainables.syn1neg.npy
│ └── doc2vec_wiki_d300_n5_w8_mc50_t12_e10_dbow.model.wv.vectors.npy
└── memes
├── memes_data.tsv
└── memes_reference_data.tsv
- CUDA:
conda install -c anaconda cudatoolkit=11.3
- cudnn
conda install -c anaconda cudnn=8.2.1=cuda11.3_0