Hi there,
Hope everyone is doing alright.
So I have a ckpt file ( checkpoint file) containing these informations:
OrderedDict([('model.encoder.0.conv.0.weight',
tensor([[[[ 1.2614e-01, 1.1951e-01, 1.6265e-01],
[ 2.1022e-01, 3.4994e-01, 5.9035e-01],
[ 1.3048e-01, 3.1809e-01, 4.4730e-01]],
[[-2.4336e-01, -1.3665e-01, -1.4839e-01],
[-1.2742e-01, -1.7422e-01, -6.1454e-02],
[-2.0389e-01, -1.4220e-01, -1.5854e-01]],
[[-5.5171e-02, -1.1455e-02, -3.8867e-02],
[ 1.5383e-02, 2.0011e-02, 1.7529e-01],
[-4.2312e-02, -9.9361e-03, 3.7781e-02]],
...,
I would like to know:
- From what I have understand, checkpoints contain the model weights (I am using a pre-trained model)?
But my model contains more layers than the ones found in the checkpoint file. - How do I manipulate this weights found in the checkpoint file?
Thanks
2)