Added examples

README.md

@@ -1,2 +1,21 @@
 # multilearn
-Pytorch wrapper for multi-task learning
+Pytorch wrapper for multi-task learning. The GitHub repo can be found in [here](https://github.com/leschultz/multilearn.git).
+
+## Coding Style
+
+Python scripts follow PEP 8 guidelines. A usefull tool to use to check a coding style is flake8.
+
+```
+flake8 <script>
+```
+
+## Authors
+
+### Graduate Students
+* **Lane Schultz** - *Main Contributer* - [leschultz](https://github.com/leschultz)
+
+## Acknowledgments
+
+* The [Computational Materials Group (CMG)](https://matmodel.engr.wisc.edu/) at the University of Wisconsin - Madison
+* Professor Dane Morgan [ddmorgan](https://github.com/ddmorgan) and Dr. Ryan Jacobs [rjacobs914](https://github.com/rjacobs914) for computational material science guidence
+

examples/materials/asr/fit.py

@@ -0,0 +1,39 @@
+from sklearn.preprocessing import StandardScaler
+from multilearn import datasets, models, utils
+from torch import optim, nn
+
+
+def main():
+
+    save_dir = 'outputs'
+    lr = 1e-4
+    batch_size = 32
+    n_epochs = 1000
+    tasks = ['asr']
+
+    # Data
+    X, y = datasets.load(tasks)
+    data = datasets.splitter(X, y, tasks, train_size=1)
+
+    for k, v in data.items():
+        data[k]['scaler'] = StandardScaler()
+        data[k]['loss'] = nn.L1Loss()
+
+    model = models.MultiNet(tasks=tasks, input_arch={500: 1})
+    optimizer = optim.Adam
+
+    out = utils.train(
+                      model,
+                      optimizer,
+                      data,
+                      n_epochs=n_epochs,
+                      batch_size=batch_size,
+                      lr=lr,
+                      save_dir=save_dir,
+                      )
+
+    print(out['df_loss'])
+
+
+if __name__ == '__main__':
+    main()

examples/materials/asr/run.sh

@@ -1,2 +1,2 @@
-export PYTHONPATH=../../:$PYTHONPATH
-python3 run.py
+export PYTHONPATH=../../../src/:$PYTHONPATH
+python3 fit.py

examples/materials/combined/fit.py

@@ -0,0 +1,39 @@
+from sklearn.preprocessing import StandardScaler
+from multilearn import datasets, models, utils
+from torch import optim, nn
+
+
+def main():
+
+    save_dir = 'outputs'
+    lr = 1e-4
+    batch_size = 32
+    n_epochs = 1000
+    tasks = ['asr', 'opband', 'stability']
+
+    # Data
+    X, y = datasets.load(tasks)
+    data = datasets.splitter(X, y, tasks, train_size=1)
+
+    for k, v in data.items():
+        data[k]['scaler'] = StandardScaler()
+        data[k]['loss'] = nn.L1Loss()
+
+    model = models.MultiNet(tasks=tasks, input_arch={500: 1})
+    optimizer = optim.Adam
+
+    out = utils.train(
+                      model,
+                      optimizer,
+                      data,
+                      n_epochs=n_epochs,
+                      batch_size=batch_size,
+                      lr=lr,
+                      save_dir=save_dir,
+                      )
+
+    print(out['df_loss'])
+
+
+if __name__ == '__main__':
+    main()

examples/materials/combined/run.sh

@@ -1,2 +1,2 @@
-export PYTHONPATH=../../:$PYTHONPATH
-python3 run.py
+export PYTHONPATH=../../../src/:$PYTHONPATH
+python3 fit.py