Training Summary

Concise steps to train the models, assuming the docker environment is set up and running in the terminal.

Training Dataset Preprocessing

In the terminal, first type:

cd <path_to_SIT_FUSE>/SIT_FUSE/sit_fuse/datasets/

An example of the full path would be:

app/SIT_FUSE/sit_fuse/datasets/

The path and steps to get to the specified folder will depend on the current directory of the terminal.

If frequently getting a "No such file or directory" error, try:

ls

This will allow viewing of the directory contents and help navigation to the desired directory.

Next, type:

python3 sf_dataset.py -y ../config/<folder>/<yaml_file>
# Replace the example yaml file with the desired yaml file and update the folder accordingly

An example of the path and file would be:

../config/model/gk2a_test_dbn_multi_layer_pl.yaml

Training Command

First, type:

cd <path_to_SIT_FUSE>/SIT_FUSE/sit_fuse/train/
# Same process as above

However, if already in the datasets directory, type:

cd ..
cd train
# Do this if already in SIT_FUSE/sit_fuse/datasets/

Then, type:

python3 pretrain_encoder.py -y <path_to_yaml>
# Use the same yaml file and path as above

python3 finetune_dc_mlp_head.py -y <path_to_yaml>

python3 train_heirarchichal_deep_cluster.py -y <path_to_yaml>

Inference Command

First, type:

cd <path_to_SIT_FUSE>/SIT_FUSE/sit_fuse/inference/
# Same process as above

However, if already in the train directory, type:

cd ..
cd inference
# Do this if already in SIT_FUSE/sit_fuse/train/

Then, type:

python3 generate_output.py -y <path_to_yaml>
# Use the same yaml file and path as above

Summary

An example training and the steps to implement it:

Training Dataset Preprocessing

cd /app/SIT_FUSE/sit_fuse/datasets/

python3 sf_dataset.py -y ../config/model/gk2a_test_dbn_multi_layer_pl.yaml

Training Command

cd /train

python3 pretrain_encoder.py -y ../config/model/gk2a_test_dbn_multi_layer_pl.yaml

python3 finetune_dc_mlp_head.py -y ../config/model/gk2a_test_dbn_multi_layer_pl.yaml

python3 train_heirarchichal_deep_cluster.py -y ../config/model/gk2a_test_dbn_multi_layer_pl.yaml

Inference Command

cd /inference

python3 generate_output.py -y ../config/model/gk2a_test_dbn_multi_layer_pl.yaml