Saving my validation notes here in case it helps others run this model, or for if I need to run the model in the future :)
* Create a `p2.xlarge` instance using the following AMI: `Deep Learning AMI (Amazon Linux 2) Version 27.0`
* `pip install synapseclient`
* `docker login docker.synapse.org`
* `docker pull docker.synapse.org/syn21560898/ctd2_atom:v1`
* `mkdir input`
* `cd input`
* `synapse get syn21321426` (get the template file as input)
* `cd ~`
* `mkdir output`
* test without training: `docker run -it --rm -v ${PWD}/input:/workspace/input -v ${PWD}/output:/workspace/output -e GPU_NO="0" -e LOAD_MODEL="load" --runtime=nvidia docker.synapse.org/syn21560898/ctd2_atom:v1`
Will run through several processes, and will output predictions.csv in the `output` dir.
Saved output to main challenge project [here](syn21769349).