Merge branch 'master' into attn_update

Author: rwightman

tests/test_models.py

@@ -17,7 +17,7 @@
 # transformer models don't support many of the spatial / feature based model functionalities
 NON_STD_FILTERS = [
     'vit_*', 'tnt_*', 'pit_*', 'swin_*', 'coat_*', 'cait_*', '*mixer_*', 'gmlp_*', 'resmlp_*', 'twins_*',
-    'convit_*', 'levit*', 'visformer*', 'deit*', 'jx_nest_*', 'nest_*', 'xcit_*']
+    'convit_*', 'levit*', 'visformer*', 'deit*', 'jx_nest_*', 'nest_*', 'xcit_*', 'crossvit_*', 'beit_*']
 NUM_NON_STD = len(NON_STD_FILTERS)
 
 # exclude models that cause specific test failures
@@ -188,23 +188,22 @@ def test_model_default_cfgs_non_std(model_name, batch_size):
 
     input_tensor = torch.randn((batch_size, *input_size))
 
-    # test forward_features (always unpooled)
     outputs = model.forward_features(input_tensor)
-    if isinstance(outputs, tuple):
+    if isinstance(outputs, (tuple, list)):
         outputs = outputs[0]
     assert outputs.shape[1] == model.num_features
 
     # test forward after deleting the classifier, output should be poooled, size(-1) == model.num_features
     model.reset_classifier(0)
     outputs = model.forward(input_tensor)
-    if isinstance(outputs, tuple):
+    if isinstance(outputs,  (tuple, list)):
         outputs = outputs[0]
     assert len(outputs.shape) == 2
     assert outputs.shape[1] == model.num_features
 
     model = create_model(model_name, pretrained=False, num_classes=0).eval()
     outputs = model.forward(input_tensor)
-    if isinstance(outputs, tuple):
+    if isinstance(outputs, (tuple, list)):
         outputs = outputs[0]
     assert len(outputs.shape) == 2
     assert outputs.shape[1] == model.num_features

tests/test_optim.py

@@ -319,10 +319,10 @@ def test_sgd(optimizer):
     #      lambda opt: ReduceLROnPlateau(opt)]
     # )
     _test_basic_cases(
-        lambda weight, bias: create_optimizer_v2([weight, bias], optimizer, lr=1e-3, momentum=1)
+        lambda weight, bias: create_optimizer_v2([weight, bias], optimizer, lr=3e-3, momentum=1)
     )
     _test_basic_cases(
-        lambda weight, bias: create_optimizer_v2([weight, bias], optimizer, lr=1e-3, momentum=1, weight_decay=.1)
+        lambda weight, bias: create_optimizer_v2([weight, bias], optimizer, lr=3e-3, momentum=1, weight_decay=.1)
     )
     _test_rosenbrock(
         lambda params: create_optimizer_v2(params, optimizer, lr=1e-3)

timm/models/__init__.py

@@ -1,8 +1,10 @@
+from .beit import *
 from .byoanet import *
 from .byobnet import *
 from .cait import *
 from .coat import *
 from .convit import *
+from .crossvit import *
 from .cspnet import *
 from .densenet import *
 from .dla import *
@@ -36,6 +38,7 @@
 from .swin_transformer import *
 from .tnt import *
 from .tresnet import *
+from .twins import *
 from .vgg import *
 from .visformer import *
 from .vision_transformer import *
@@ -44,7 +47,6 @@
 from .xception import *
 from .xception_aligned import *
 from .xcit import *
-from .twins import *
 
 from .factory import create_model, split_model_name, safe_model_name
 from .helpers import load_checkpoint, resume_checkpoint, model_parameters