https://github.com/urish/nn-function-generator

Experimenting with automatic generation of TS function bodies using ANN models
https://github.com/urish/nn-function-generator

bigquery tensorflow tsquery typescript

Last synced: 2 days ago
JSON representation

Experimenting with automatic generation of TS function bodies using ANN models

• Host: GitHub
• URL: https://github.com/urish/nn-function-generator
• Owner: urish
• Created: 2019-01-10T22:07:32.000Z (almost 6 years ago)
• Default Branch: master
• Last Pushed: 2021-05-10T11:54:59.000Z (over 3 years ago)
• Last Synced: 2024-05-10T03:43:32.195Z (6 months ago)
• Topics: bigquery, tensorflow, tsquery, typescript
• Language: TypeScript
• Homepage: https://nn-functions-playground-myeqrox3aa-uc.a.run.app/
• Size: 180 KB
• Stars: 8
• Watchers: 5
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# nn-function-generator

Experimenting with automatic generation of TS function bodies using ANN models

## Setup

1. Download `typescript-all-functions.json.gz` and put it inside the `data` directory

2. Run `yarn dataset` to prepare the dataset for training

3. Run `cd model && python train.py` to train the model

## Running the playground

1. Run `cd model && python test-server.py --run run_001`

2. Run `yarn start`

3. Go to `http://localhost:3003/` and have fun!

Alternatively, you can use docker and generate an image for the playground app using the

provided [Dockerfile](Dockerfile):

```bash

docker build -t nn-function-gen .

docker -it -v /path/to/dataset:/app/data -v /path/to/model/runs:/app/runs -p 3003:3003 run nn-function-gen 

```

Remember to replace `/path/to/dataset` and `/path/to/model/runs` with the appropriate paths for your machine. The image is configured to load the model from the first run (`run_001`) by default.

After starting the container, you can go to [http://localhost:3003](http://localhost:3003/) to launch the playground.

## Training on Google Cloud TPU

```

ctpu up -preemptible -machine-type n1-highmem-16	

virtualenv ngvikings

source ngvikings/bin/activate

cd model

pip install -r requirements.txt

python train.py --tpu

```