gistfile1.txt

index ede2865..de5eb9f 100755
--- a/examples/maml-omniglot.py
+++ b/examples/maml-omniglot.py
@@ -30,6 +30,7 @@ import higher

 from omniglot_loaders import OmniglotNShot

+
 def main():
     argparser = argparse.ArgumentParser()
     argparser.add_argument('--n_way', type=int, help='n way', default=5)
@@ -58,23 +59,22 @@ def main():
     # Before higher, models could *not* be created like this
     # and the parameters needed to be manually updated and copied
     # for the updates.
+
     net = nn.Sequential(
-        nn.Conv2d(1, 64, 3, 2),
-        nn.ReLU(inplace=True),
-        nn.BatchNorm2d(64),
-        nn.Conv2d(64, 64, 3, 2),
+        nn.Conv2d(1, 64, 3),
+        nn.BatchNorm2d(64, momentum=1, affine=True),
         nn.ReLU(inplace=True),
-        nn.BatchNorm2d(64),
-        nn.Conv2d(64, 64, 3, 2),
+        nn.MaxPool2d(2, 2),
+        nn.Conv2d(64, 64, 3),
+        nn.BatchNorm2d(64, momentum=1, affine=True),
         nn.ReLU(inplace=True),
-        nn.BatchNorm2d(64),
-        nn.Conv2d(64, 64, 2, 1),
+        nn.MaxPool2d(2, 2),
+        nn.Conv2d(64, 64, 3),
+        nn.BatchNorm2d(64, momentum=1, affine=True),
         nn.ReLU(inplace=True),
-        nn.BatchNorm2d(64),
+        nn.MaxPool2d(2,2),
         Flatten(),
-        nn.Linear(64, args.n_way)
-    ).to(device)
-
+        nn.Linear(64, args.n_way)).to(device)
     # We will use Adam to (meta-)optimize the initial parameters
     # to be adapted.
     meta_opt = optim.Adam(net.parameters(), lr=1e-3)
@@ -90,6 +90,7 @@ def train(db, net, device, meta_opt, epoch, log):
     net.train()
     n_train_iter = db.x_train.shape[0] // db.batchsz

+
     for batch_idx in range(n_train_iter):
         # Sample a batch of support and query images and labels.
         x_spt, y_spt, x_qry, y_qry = db.next()
@@ -107,7 +108,7 @@ def train(db, net, device, meta_opt, epoch, log):
         # Initialize the inner optimizer to adapt the parameters to
         # the support set.
         n_inner_iter = 1
-        inner_opt = torch.optim.SGD(net.parameters(), lr=4e-1)
+        inner_opt = torch.optim.SGD(net.parameters(), lr=1e-1)

         qry_losses = []
         qry_accs = []
@@ -167,6 +168,7 @@ def test(db, net, device, epoch, log):

     qry_losses = []
     qry_accs = []
+

     for batch_idx in range(n_test_iter):
         x_spt, y_spt, x_qry, y_qry = db.next('test')
@@ -180,8 +182,8 @@ def test(db, net, device, epoch, log):

         # TODO: Maybe pull this out into a separate module so it
         # doesn't have to be duplicated between `train` and `test`?
-        n_inner_iter = 3
-        inner_opt = torch.optim.SGD(net.parameters(), lr=4e-1)
+        n_inner_iter = 5
+        inner_opt = torch.optim.SGD(net.parameters(), lr=1e-1)

         for i in range(task_num):
             with higher.innerloop_ctx(net, inner_opt) as (fnet, diffopt):
@@ -195,7 +197,7 @@ def test(db, net, device, epoch, log):

                 # The query loss and acc induced by these parameters.
                 qry_logits = fnet(x_qry[i]).detach()
-                qry_loss = F.cross_entropy(qry_logits, y_qry[i], reduction='none')
+                qry_loss = F.cross_entropy(qry_logits, y_qry[i], reduction='none').detach()
                 qry_losses.append(qry_loss)
                 qry_accs.append(
                     qry_logits.argmax(dim=1) == y_qry[i]