Closed zhang0931 closed 5 years ago
Try running it with --ignore-missing-variables
. Be aware that this randomly initializes the variables not available in the checkpoint you provide (the variables listed in the error message).
hi, I execute the following: python3 trainer_views.py --output=everything_generate/output_dir --data=/media/lab210/Work/WH_HOME/SubjectDownload/Project/DATASETS/datasets/RAP/RAP_dataset --dataset-name=rap --batch-size=16 --num-epochs=100 --network-name=resnet_v1_50_views --initial-checkpoint=/media/lab210/Work/WH_HOME/SubjectDownload/Project/reid/pose-sensitive-embedding/everything_generate/checkpoint/resnet_v1_50.ckpt --checkpoint-exclude-scopes=everything_generate/resnet_v1_50/logits --trainable-scopes=resnet_v1_50/3Views --ignore-missing-variables
but get the error:
The 'resnet_v1_50.ckpt' was downloaded from 'https://github.com/tensorflow/models/tree/master/research/slim'
And another question:
--checkpoint-exclude-scopes=everything_generate/resnet_v1_50/logits
where is the logits?
try --checkpoint-exclude-scopes= resnet_v1_50/logits i.e. with only the network name, dont need to append the path
Thank you for your help!
在 2018-11-01 02:21:46,ssarfraz notifications@github.com 写道:
try --checkpoint-exclude-scopes= resnet_v1_50/logits i.e. with only the network name, dont need to append the path
— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.
hi,
Thank the author, you job is so great. I successfully trained once, but there was a problem when I trained again .The problem is shown in the screenshot I execute the following: sudo python3 trainer_views.py --output=./output/Views_mode_new --data=/data/zhangli/datasets/RAP_dataset --dataset-name=rap --batch-size=16 --num-epochs=100 --network-name=resnet_v1_50_views --initial-checkpoint=/data/zhangli/pose-sensitive-embedding-master/checkpoint_dir/resnet_v1_50.ckpt --checkpoint-exclude-scopes=resnet_v1_50/logits --trainable-scopes=resnet_v1_50/3Views