dvc_commands

DOC : https://dvc.org/doc/get-started/visualize


PIPELINE STAGES ;

1. GET DATA - download data from gdrive
2. ANALYSIS - create figures in order to create reports
3. FEATURES - preprocessing stage + train/test/valid split
4. TRAINING -

# DATA

dvc run -f .dvc_files/download_features.dvc \
        -d src/data/download_dataset.py \
        -o data/raw \
        python src/data/download_dataset.py


# ANALYSIS

dvc run -f .dvc_files/analyze_data.dvc \
        -d data/raw \
        -d src/visualization/visualize.py \
        python src/visualization/visualize.py


# FEATURES

dvc run -f .dvc_files/build_features.dvc \
          -d src/features/build_features.py -d data/raw \
          -o data/preprocessed/X_train.npy \
	  -o data/preprocessed/X_test.npy \
	  -o data/preprocessed/X_valid.npy \
          -o data/preprocessed/y_train.npy \
	  -o data/preprocessed/y_test.npy \
	  -o data/preprocessed/y_valid.npy \
          python src/features/build_features.py

# TRAINING

dvc run -f .dvc_files/train_models.dvc \
          -d src/models/train_model.py -d data/preprocessed \
       	  -o models/VAE.hdf5 \
       	  -o reports/figures/learning_curve.png \
       	  -M models/train_metrics.txt\
          python src/models/train_model.py -e 100

# EVALUATION

dvc run -f .dvc_files/test_model.dvc \
          -d src/models/test_model.py \
          -d data/preprocessed \
          -d models/VAE.hdf5 \
          -o reports/figures/visual_eval.png \
          -M models/test_metrics.txt \
          python src/models/test_model.py

# VISUALIZATION 

# Visualize stages 
dvc pipeline show --ascii .dvc_files/test.dvc

# Visualize commands
dvc pipeline show --ascii .dvc_files/test.dvc --commands

# Visualize outputs
dvc pipeline show --ascii .dvc_files/test.dvc --outs

# REPRODUCE EXPERIMENT

dvc repro .dvc_files/test.dvc