Skip to content

Latest commit

 

History

History
19 lines (14 loc) · 873 Bytes

README.md

File metadata and controls

19 lines (14 loc) · 873 Bytes

Spikefinder submission

Reproduces my final submission to Spikefinder.

Steps to retrain model from scratch:

  • Clone this repo
  • cd into spikefinder_submission and clone mixture density network model via git clone https://github.com/patrickmineault/write-rnn-tensorflow.git write_rnn_tensorflow
  • Train mixture density network latent features via python train_and_eval_mdn.py --operation=train
  • Eval mixture density network latent features via python train_and_eval_mdn.py --operation=eval
  • Fit core model and refine for each of the first 5 recording sets by running ./train_and_refine.sh
  • Eval for the unrefined and the refined model via ./eval_all.sh

Steps to predict using pretrained model via Docker.

  • Clone this repo.
  • sudo docker build -t patrickmineault:spikefinder .
  • sudo docker run -it patrickmineault:spikefinder