New FX test strategy, filter based on param count

tests/test_models.py

@@ -399,8 +399,10 @@ def test_model_backward_fx(model_name, batch_size):
         pytest.skip("Fixed input size model > limit.")
 
     model = create_model(model_name, pretrained=False, num_classes=42)
-    num_params = sum([x.numel() for x in model.parameters()])
     model.train()
+    num_params = sum([x.numel() for x in model.parameters()])
+    if 'GITHUB_ACTIONS' in os.environ and num_params > 100e6:
+        pytest.skip("Skipping FX backward test on model with more than 100M params.")
 
     model = _create_fx_model(model, train=True)
     outputs = tuple(model(torch.randn((batch_size, *input_size))).values())