- A demo can be found on Youtube
- Link to devpost here
- [NSFW] Release all you can say via Swagger under /predict
biLSTM
trained oniMDB
for sentiment analysistorchtext
was built frommaster
, which can be found here
make help
for further instructions
- prepare pretrained embeddings do
make init
- To package the model to bento do
make build
- to run the docker container do
make run
- endpoint can be accessed through
POST
request vialocalhost:5000
curl -X POST 0.0.0.0:5000/predict -H "accept: */*" -H "Content-Type: application/json" \
-d "{\"text\":\"I hate you this is the worst experience I have ever seen\"}"
# > 0.2898484170436859
- do
git clone --recurse-submodules https://github.com/MLH-Fellowship/1.0.5-DiscoBot && cd torchtext
git submodules update --init --recursive && python setup.py clean install
to build torchtext from source (currently at 0.8.0a0+8dc2125)- to prep the pretrained embedding run
make prep
, otherwise if you want to train the model parsesARGS=--train
, like so:
- to prep the pretrained embedding run
# This will train the model
make ARGS=--train prep