v0.0.28

mapk can handle devices

Author: bartzbeielstein

pyproject.toml

@@ -7,7 +7,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "spotPython"
-version = "0.2.26"
+version = "0.2.28"
 authors = [
   { name="T. Bartz-Beielstein", email="tbb@bartzundbartz.de" }
 ]

src/spotPython/torch/mapk.py

@@ -8,6 +8,7 @@ class MAPK(Metric):
     Args:
         k: Number of predictions to consider
         dist_sync_on_step: Whether to sync the output across all GPUs
+        device: Device to use for the computation
     Example:
         >>> from torchmetrics import MAPK
         >>> target = torch.tensor([0, 1, 2, 3])
@@ -30,8 +31,8 @@ class MAPK(Metric):
         >>> print(result) # tensor(0.37500)
     """
 
-    def __init__(self, k=3, dist_sync_on_step=False):
-        super().__init__(dist_sync_on_step=dist_sync_on_step)
+    def __init__(self, k=3, dist_sync_on_step=False, device=None):
+        super().__init__(dist_sync_on_step=dist_sync_on_step, device=device)
         self.k = k
         self.add_state("actual", default=[], dist_reduce_fx="cat")
         self.add_state("predicted", default=[], dist_reduce_fx="cat")

src/spotPython/torch/traintest.py

@@ -113,7 +113,6 @@ def evaluate_cv(
                 print("We will use", torch.cuda.device_count(), "GPUs!")
                 net = nn.DataParallel(net)
         net.to(device)
-        metric.to(device)
         optimizer = optimizer_handler(
             optimizer_name=optimizer_instance,
             params=net.parameters(),
@@ -220,7 +219,6 @@ def evaluate_hold_out(
                 print("We will use", torch.cuda.device_count(), "GPUs!")
                 net = nn.DataParallel(net)
         net.to(device)
-        metric.to(device)
         # loss_function = nn.CrossEntropyLoss()
         # TODO: optimizer = optim.Adam(net.parameters(), lr=lr)
         # optimizer = optim.SGD(net.parameters(), lr=lr, momentum=0.9)
@@ -266,7 +264,6 @@ def evaluate_hold_out(
             metric_name = "Metric"
             if metric is not None:
                 metric_name = type(metric).__name__
-                print(f"{metric_name} value on hold-out data: {metric_val}")
             if writer is not None:
                 writer.add_scalars(
                     "evaluate_hold_out: Train & Val Loss and Val Metric" + writerId,