Even for those already acquainted with ML and popular ML libraries it is often a frustrating process to setup a ML model. One is led down a winding path of dependencies, ReadMes, and stale model downloads. Since many of these models can be thought as simple functions and there already exist many excellent pre-trained models, we thought it was silly that people have been wasting so much time setting up the same models over and over again. What if we could abstract the setup, and allow people to leverage model inference with one line of code?
What it does
We built a python library (leveraging paddlepaddle) that allows people to leverage word2vec, inception image classification, sentiment classification, machine translation, digit recognition, or object recognition, with one line of code. Since setup is extracted away, it's possible to quickly use and iterate upon various models.
How we built it
We built a python library which supports various ML models running on PaddlePaddle. We have code which checks and gathers dependencies when running a model.
Challenges we ran into
There are many dependency nuances and there were subtleties restarting docker images for some of the models we used.
Accomplishments that we're proud of
Log in or sign up for Devpost to join the conversation.