Evaluation the Experimental Design of Land Cover Classification Model
- Create a new conda environment and activate it
conda create --name cs294 conda activate cs294 conda install pip
- Install requirement packages
You might want to install the cuda version of pytorch again using the cmd here if you want to use gpu for pytorch
pip install -r requirements.txt
-
run on your own laptop
python -m model_baseline.resnet-eurosat
-
run on bridge2
# login in to interactive node salloc -N 1 -n 1 -p GPU-shared --gres=gpu:1 -q interactive -t 01:00:00 # load pytorch environment singularity shell /ocean/containers/ngc/pytorch/pytorch_latest.sif # run code python -m model_baseline.resnet-eurosat
-
run on bridges-2 job
change fc_nodes in bridge_job.sh then run
sbatch bridge_job.sh
-
Download the dataset and try start resnet base training
python -m model_baseline.resnet-eurosat
-
Generate csv data from RGB photo and resnet output
python -m scripts.image-to-csv python -m scripts.resnet-to-csv
-
run brainome
python -m brainome login python -m brainome -vv -target class -o resnet-nn.py -f NN ./data/resnet-output.csv
-
Training model with different hiden layer nodes
for i in {1..15} do python -m model_baseline.mlp-eurosat --fc_nodes $i done
-
run notebook under ./scripts/data_mec.ipynb to get dataset mec, Capacity Progression and plots