Refactored slightly to allow other experiments

symbolic_nn_tests/__main__.py

@@ -1,8 +1,8 @@
 LEARNING_RATE = 10e-5
 
 
-def run_test(loss_func, version):
+def qmnist_test(loss_func, version):
-    from .model import main as test_model
+    from .qmnist.model import main as test_model
     from lightning.pytorch.loggers import TensorBoardLogger
     from lightning.pytorch.loggers import WandbLogger
     import wandb
@@ -12,26 +12,34 @@ def run_test(loss_func, version):
         name="logs/comparison",
         version=version,
     )
-    wandb_logger = WandbLogger(
+    # wandb_logger = WandbLogger(
-        project="Symbolic_NN_Tests",
+    #     project="Symbolic_NN_Tests",
-        name=version,
+    #     name=version,
-        dir="wandb",
+    #     dir="wandb",
-    )
+    # )
-    logger = [tb_logger, wandb_logger]
+    logger = [
+        tb_logger,
+    ]  # wandb_logger]
     test_model(logger=logger, loss_func=loss_func, lr=LEARNING_RATE)
     wandb.finish()
 
 
-def main():
+def qmnist_experiment():
-    from . import semantic_loss
+    from .qmnist import semantic_loss
     from torch import nn
 
-    run_test(nn.functional.cross_entropy, "cross_entropy")
+    qmnist_test(nn.functional.cross_entropy, "cross_entropy")
-    run_test(semantic_loss.similarity_cross_entropy, "similarity_cross_entropy")
+    qmnist_test(semantic_loss.similarity_cross_entropy, "similarity_cross_entropy")
-    run_test(semantic_loss.hasline_cross_entropy, "hasline_cross_entropy")
+    qmnist_test(semantic_loss.hasline_cross_entropy, "hasline_cross_entropy")
-    run_test(semantic_loss.hasloop_cross_entropy, "hasloop_cross_entropy")
+    qmnist_test(semantic_loss.hasloop_cross_entropy, "hasloop_cross_entropy")
-    run_test(semantic_loss.multisemantic_cross_entropy, "multisemantic_cross_entropy")
+    qmnist_test(
-    run_test(semantic_loss.garbage_cross_entropy, "garbage_cross_entropy")
+        semantic_loss.multisemantic_cross_entropy, "multisemantic_cross_entropy"
+    )
+    qmnist_test(semantic_loss.garbage_cross_entropy, "garbage_cross_entropy")
+
+
+def main():
+    qmnist_experiment()
 
 
 if __name__ == "__main__":

symbolic_nn_tests/qmnist/model.py

@@ -14,7 +14,7 @@ model = nn.Sequential(
 def get_singleton_dataset():
     from torchvision.datasets import QMNIST
 
-    from .dataloader import get_dataset
+    from symbolic_nn_tests.dataloader import get_dataset
 
     return get_dataset(dataset=QMNIST)
 
@@ -22,7 +22,7 @@ def get_singleton_dataset():
 def main(loss_func=nn.functional.cross_entropy, logger=None, **kwargs):
     import lightning as L
 
-    from .train import TrainingWrapper
+    from symbolic_nn_tests.train import TrainingWrapper
 
     if logger is None:
         from lightning.pytorch.loggers import TensorBoardLogger