Hi! I load checkpoint from model with head size = 1599 to same model with head size = 59.
Set strict=False, but got the error:
Traceback (most recent call last):
File "/home/jovyan/murtazin/NLA-SLR/train.py", line 35, in <module>
model = VideoClassificationModel.load_from_checkpoint(ckpth_path,
File "/home/user/conda/envs/nla/lib/python3.9/site-packages/lightning/pytorch/core/module.py", line 1537, in load_from_checkpoint
loaded = _load_from_checkpoint(
File "/home/user/conda/envs/nla/lib/python3.9/site-packages/lightning/pytorch/core/saving.py", line 91, in _load_from_checkpoint
model = _load_state(cls, checkpoint, strict=strict, **kwargs)
File "/home/user/conda/envs/nla/lib/python3.9/site-packages/lightning/pytorch/core/saving.py", line 157, in _load_state
keys = obj.load_state_dict(checkpoint["state_dict"], strict=strict)
File "/home/user/conda/envs/nla/lib/python3.9/site-packages/torch/nn/modules/module.py", line 2041, in load_state_dict
raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
RuntimeError: Error(s) in loading state_dict for VideoClassificationModel:
size mismatch for model.model.classifier.1.weight: copying a param with shape torch.Size([1599, 1024, 1, 1, 1]) from checkpoint, the shape in current model is torch.Size([59, 1024, 1, 1, 1]).
size mismatch for model.model.classifier.1.bias: copying a param with shape torch.Size([1599]) from checkpoint, the shape in current model is torch.Size([59]).
My code:
num_classes = 59
video_model = S3D(num_classes=num_classes)
metrics = Accuracy(task="multiclass", num_classes=num_classes)
ckpth_path = 'lightning_logs/version_1/checkpoints/epoch=epoch=189-step=step=357960-val_acc=val_acc=0.66.ckpt'
print(ckpth_path)
model = VideoClassificationModel.load_from_checkpoint(ckpth_path,
model=video_model, metrics=metrics, strict=False,
)