On the Parameter Selection Problem of 'pretrained' in Creating Models
xuanRyc opened this issue · comments
I want to use a pretrained model。when I follow this code
'from models.t2t_vit import *
from utils import load_for_transfer_learning
create model
model = t2t_vit_14()
load the pretrained weights
load_for_transfer_learning(model, /path/to/pretrained/weights, use_ema=True, strict=False, num_classes=1000)',
only by turning 't2t_vit_t_14()'Setting the parameter 'pretrained' to False is necessary for normal operation, otherwise an error will be reported,just like'RuntimeError: Invalid pretrained config, cannot load weights. Use pretrained=False
for random init.'
I don't know if I actually used the parameters of the pretrained model