holyseven / PSPNet-TF-Reproduce

Training PSPNet in Tensorflow. Reproduce the performance from the paper.
MIT License
125 stars 30 forks source link

weight load error #27

Closed JankYang12 closed 4 years ago

JankYang12 commented 5 years ago

Finetuning Process: not import resnet_v1_50/block3/unit_7/weights:0 > < Finetuning Process: not import resnet_v1_50/block3/unit_7/BatchNorm/beta:0 > < Finetuning Process: not import resnet_v1_50/block3/unit_7/BatchNorm/gamma:0 > < Finetuning Process: not import resnet_v1_50/aux_logits/weights:0 > < Finetuning Process: not import resnet_v1_50/aux_logits/biases:0 > < Finetuning Process: not import resnet_v1_50/psp/pool6/weights:0 > < Finetuning Process: not import resnet_v1_50/psp/pool6/BatchNorm/beta:0 > < Finetuning Process: not import resnet_v1_50/psp/pool6/BatchNorm/gamma:0 > < Finetuning Process: not import resnet_v1_50/psp/pool3/weights:0 > < Finetuning Process: not import resnet_v1_50/psp/pool3/BatchNorm/beta:0 > < Finetuning Process: not import resnet_v1_50/psp/pool3/BatchNorm/gamma:0 > < Finetuning Process: not import resnet_v1_50/psp/pool2/weights:0 > < Finetuning Process: not import resnet_v1_50/psp/pool2/BatchNorm/beta:0 > < Finetuning Process: not import resnet_v1_50/psp/pool2/BatchNorm/gamma:0 > < Finetuning Process: not import resnet_v1_50/psp/pool1/weights:0 > < Finetuning Process: not import resnet_v1_50/psp/pool1/BatchNorm/beta:0 > < Finetuning Process: not import resnet_v1_50/psp/pool1/BatchNorm/gamma:0 > < Finetuning Process: not import resnet_v1_50/block4/unit_4/weights:0 > < Finetuning Process: not import resnet_v1_50/block4/unit_4/BatchNorm/beta:0 > < Finetuning Process: not import resnet_v1_50/block4/unit_4/BatchNorm/gamma:0 > < Finetuning Process: not import resnet_v1_50/logits/weights:0 > < Finetuning Process: not import resnet_v1_50/logits/biases:0 > < Succesfully loaded fine-tune model from ./z_pretrained_weights/resnet_v1_50.ckpt. >

< training process begins >

loss or weight norm is nan. Training Stopped!

JankYang12 commented 5 years ago

Oh, I solved it. just because my batch_size is too small