Could not find the number of trainable parameters

[NickNtamp]

Hello, I tried to custom train detectron2 on my dataset using faster_rcnn_X_101_32x8d_FPN_3x.yaml model as backbone for object detection.
I am curious of finding the exact number of trainable parameters I trained using the pre-trained model mentioned above.
I tried to print them using the below code block for total/trainable parameters but I believe that actually I trained much less parameters and the numbers I get are referring to the whole model's parameters:

from torch.nn.utils import parameters_to_vector
print(parameters_to_vector(predictor.model.parameters()).numel())
pytorch_total_params = sum(p.numel() for p in predictor.model.parameters() if p.requires_grad)
print(pytorch_total_params)

104384730
103959578

Can you please help me on this -to determine the exact number of parameters I trained? Thank you.