Question regarding Video Classifier.

[alitirmizi6200]

So, I was following our brother Innat's work. His notebooks regarding video swin transformer. I am fairly new to python. So untill te migration happens of Innat's keras_cs repo to Keras_team repo, Code was executing just fine but know it shows error's like model.parameters() not found or model.train() not found.
I am sharing relevant code from notebook.

`import keras
from keras import ops
from keras_cv.models import VideoSwinBackbone
from keras_cv.models import VideoClassifier
keras.version # i.e 3.0.5'

def vswin_tiny():
backbone=VideoSwinBackbone(
input_shape=(32, 224, 224, 3),
embed_dim=96,
depths=[2, 2, 6, 2],
num_heads=[3, 6, 12, 24],
include_rescaling=False,
)
backbone.load_weights(
'/kaggle/input/videoswin/keras/tiny/1/videoswin_tiny_kinetics400.weights.h5'
)
backbone.trainable = False
keras_model = VideoClassifier(
backbone=backbone,
num_classes=len(class_folders),
activation=None,
pooling='avg',
)
return keras_model

model = vswin_tiny()
optimizer = torch.optim.Adam(
model.parameters(), lr=1e-3
)`
AttributeError: 'VideoClassifier' object has no attribute 'parameters'

Hope this was correct format, If I am doing anything wrong, guide me.

[innat]

Seems like you are using default backend, which is tensorflow. And it doesn't have .parameter API for the model. But if you swith to torch backend, then it should work. Either way, the following is the general methods to get model's param info.

model.summary()

n_parameters = model.count_params()
print("number of params (M): %.2f" % (n_parameters / 1.0e6))

And as mentioned, if you swith to torch backend, then the following should work also

n_parameters = sum(p.numel() for p in model.parameters() if p.requires_grad)
print("number of params (M): %.2f" % (n_parameters / 1.0e6))