問題記錄:恢復某些層引數,遇到NotFoundError: Key conv2d_168/bias not found in checkpoint
阿新 • • 發佈:2019-01-06
問題描述:對網路結構做了些修改,匯入已訓練的引數,出現NotFoundError
相關程式:
#State where your log file is at. log_dir = './log2' #State where your checkpoint file is checkpoint_file = './log/model.ckpt-300068' #Define the scopes that you want to exclude for restoration exclude = ['conv_0','conv_1','conv_2'] variables_to_restore = slim.get_variables_to_restore(exclude=exclude) #Now we create a saver function that actually restores the variables from a checkpoint file in a sess saver = tf.train.Saver(variables_to_restore) def restore_fn(sess): return saver.restore(sess,checkpoint_file) sv = tf.train.Supervisor(logdir = log_dir,summary_op = None, init_fn = restore_fn)
報錯:
問題原因:沒有匯入引數
問題語句:
exclude = ['conv_0','conv_1','conv_2']
variables_to_restore = slim.get_variables_to_restore(exclude=exclude)
解決措施:
include = ['dense_1','dense_2','dense_3']
variables_to_restore = slim.get_variables_to_restore(include=include)