Closed AbhayKoushik closed 6 years ago
Can you print out the lists of tf.trainable_variables()
, train_vars1
and train_vars2
?
As in the screenshot above: Optimizing parameters is same as train_vars2 which happens to be empty as shown in the screenshot below which is exactly why the ValueError. Pretrained parameters is same as train_vars1 which is not an empty list.
My guest is that you use the incorrect class for this trainer.
train_vars1
should contains parameters of the two CNNs
train_vars2
should contains parameters of the Bi-LSTM
Can you please check whether you use DeepFeatureNet
instead of DeepSleepNet
?
I too think that is the problem, As you suggested I printed those values again, here they are.
trainable_variables
[<tf.Variable 'deepsleepnet/l1_conv/conv1d/weights:0' shape=(50, 1, 1, 64) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l1_conv/bn/beta:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l1_conv/bn/gamma:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l4_conv/conv1d/weights:0' shape=(8, 1, 64, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l4_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l4_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l5_conv/conv1d/weights:0' shape=(8, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l5_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l5_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l6_conv/conv1d/weights:0' shape=(8, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l6_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l6_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l9_conv/conv1d/weights:0' shape=(400, 1, 1, 64) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l9_conv/bn/beta:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l9_conv/bn/gamma:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l12_conv/conv1d/weights:0' shape=(6, 1, 64, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l12_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l12_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l13_conv/conv1d/weights:0' shape=(6, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l13_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l13_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l14_conv/conv1d/weights:0' shape=(6, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l14_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l14_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l19_softmax_linear/weights:0' shape=(3072, 5) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l19_softmax_linear/biases:0' shape=(5,) dtype=float32_ref>]
train_vars1
[<tf.Variable 'deepsleepnet/l1_conv/conv1d/weights:0' shape=(50, 1, 1, 64) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l1_conv/bn/beta:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l1_conv/bn/gamma:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l4_conv/conv1d/weights:0' shape=(8, 1, 64, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l4_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l4_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l5_conv/conv1d/weights:0' shape=(8, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l5_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l5_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l6_conv/conv1d/weights:0' shape=(8, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l6_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l6_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l9_conv/conv1d/weights:0' shape=(400, 1, 1, 64) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l9_conv/bn/beta:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l9_conv/bn/gamma:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l12_conv/conv1d/weights:0' shape=(6,pretrained_model_name = "DeepSleepNet" 1, 64, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l12_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l12_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l13_conv/conv1d/weights:0' shape=(6, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l13_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l13_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l14_conv/conv1d/weights:0' shape=(6, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l14_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l14_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l19_softmax_linear/weights:0' shape=(3072, 5) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l19_softmax_linear/biases:0' shape=(5,) dtype=float32_ref>]
train_vars2
[]
Traceback (most recent call last):
File "train.py", line 95, in <module>
tf.app.run()
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/platform/app.py", line 124, in run
_sys.exit(main(argv))
File "train.py", line 90, in main
n_epochs=FLAGS.finetune_epochs
File "train.py", line 72, in finetune
resume=FLAGS.resume
File "/home/Abhay/SleepInterfacing/DeepSleepNet/deepsleepnet/deepsleep/trainer.py", line 633, in finetune
clip_value=10.0
File "/home/Abhay/SleepInterfacing/DeepSleepNet/deepsleepnet/deepsleep/optimize.py", line 60, in adam_clipping_list_lr
apply_gradient_op = opt.apply_gradients(gvs)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/training/optimizer.py", line 493, in apply_gradients
raise ValueError("No variables provided.")
ValueError: No variables provided.>
This is using DeepFeatureNet in the finetune function of the trainer.py
pretrained_model_name = "DeepFeatureNet"
CNN Parameters ,that is, train_vars1
is available and the pre-training is complete.
Bidirectional LSTM parameters, that is, are empty.
And Next, after the convolutionally trained model is saved,
I changed:
pretrained_model_name = "DeepSleepNet"
and I have the same error but now train_vars1
is empty and train_vars2
is available.
trainable_variables
[<tf.Variable 'deepsleepnet/l1_conv/conv1d/weights:0' shape=(50, 1, 1, 64) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l1_conv/bn/beta:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l1_conv/bn/gamma:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l4_conv/conv1d/weights:0' shape=(8, 1, 64, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l4_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l4_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l5_conv/conv1d/weights:0' shape=(8, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l5_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l5_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l6_conv/conv1d/weights:0' shape=(8, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l6_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l6_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l9_conv/conv1d/weights:0' shape=(400, 1, 1, 64) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l9_conv/bn/beta:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l9_conv/bn/gamma:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l12_conv/conv1d/weights:0' shape=(6, 1, 64, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l12_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l12_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l13_conv/conv1d/weights:0' shape=(6, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l13_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l13_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l14_conv/conv1d/weights:0' shape=(6, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l14_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l14_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l19_softmax_linear/weights:0' shape=(3072, 5) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l19_softmax_linear/biases:0' shape=(5,) dtype=float32_ref>]
train_vars1
[]
train_vars2
[<tf.Variable 'deepsleepnet/l4_conv/conv1d/weights:0' shape=(8, 1, 64, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l1_conv/bn/beta:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l4_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l9_conv/bn/beta:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l5_conv/conv1d/weights:0' shape=(8, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l6_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l9_conv/conv1d/weights:0' shape=(400, 1, 1, 64) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l13_conv/conv1d/weights:0' shape=(6, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l6_conv/conv1d/weights:0' shape=(8, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l9_conv/bn/gamma:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l14_conv/conv1d/weights:0' shape=(6, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l12_conv/conv1d/weights:0' shape=(6, 1, 64, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l5_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l5_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l1_conv/conv1d/weights:0' shape=(50, 1, 1, 64) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l6_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l19_softmax_linear/weights:0' shape=(3072, 5) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l12_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l13_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l1_conv/bn/gamma:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l13_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l4_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l12_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l19_softmax_linear/biases:0' shape=(5,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l14_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l14_conv/bn/beta:0' shape=(128,) dtype=float32_ref>]
Traceback (most recent call last):
File "train.py", line 95, in <module>
tf.app.run()
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/platform/app.py", line 124, in run
_sys.exit(main(argv))
File "train.py", line 90, in main
n_epochs=FLAGS.finetune_epochs
File "train.py", line 72, in finetune
resume=FLAGS.resume
File "/home/Abhay/SleepInterfacing/DeepSleepNet/deepsleepnet/deepsleep/trainer.py", line 633, in finetune
clip_value=10.0
File "/home/Abhay/SleepInterfacing/DeepSleepNet/deepsleepnet/deepsleep/optimize.py", line 60, in adam_clipping_list_lr
apply_gradient_op = opt.apply_gradients(gvs)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/training/optimizer.py", line 493, in apply_gradients
raise ValueError("No variables provided.")
ValueError: No variables provided.
So apparently we need both together for the optimization code to run. Can you please help me now?
Can you help debugging the L597-L599 in your figure below?
https://user-images.githubusercontent.com/17530993/35916175-8a7f103e-0c2f-11e8-8686-9f375304a889.png
This can be done by printing all variables involved in these lines (before and after these lines are executed).
I am trying to identify the root cause.
Here they are
Trainable Variables:
deepsleepnet/l1_conv/conv1d/weights:0 (50, 1, 1, 64)
deepsleepnet/l1_conv/bn/beta:0 (64,)
deepsleepnet/l1_conv/bn/gamma:0 (64,)
deepsleepnet/l4_conv/conv1d/weights:0 (8, 1, 64, 128)
deepsleepnet/l4_conv/bn/beta:0 (128,)
deepsleepnet/l4_conv/bn/gamma:0 (128,)
deepsleepnet/l5_conv/conv1d/weights:0 (8, 1, 128, 128)
deepsleepnet/l5_conv/bn/beta:0 (128,)
deepsleepnet/l5_conv/bn/gamma:0 (128,)
deepsleepnet/l6_conv/conv1d/weights:0 (8, 1, 128, 128)
deepsleepnet/l6_conv/bn/beta:0 (128,)
deepsleepnet/l6_conv/bn/gamma:0 (128,)
deepsleepnet/l9_conv/conv1d/weights:0 (400, 1, 1, 64)
deepsleepnet/l9_conv/bn/beta:0 (64,)
deepsleepnet/l9_conv/bn/gamma:0 (64,)
deepsleepnet/l12_conv/conv1d/weights:0 (6, 1, 64, 128)
deepsleepnet/l12_conv/bn/beta:0 (128,)
deepsleepnet/l12_conv/bn/gamma:0 (128,)
deepsleepnet/l13_conv/conv1d/weights:0 (6, 1, 128, 128)
deepsleepnet/l13_conv/bn/beta:0 (128,)
deepsleepnet/l13_conv/bn/gamma:0 (128,)
deepsleepnet/l14_conv/conv1d/weights:0 (6, 1, 128, 128)
deepsleepnet/l14_conv/bn/beta:0 (128,)
deepsleepnet/l14_conv/bn/gamma:0 (128,)
deepsleepnet/l19_softmax_linear/weights:0 (3072, 5)
deepsleepnet/l19_softmax_linear/biases:0 (5,)
All Variables:
deepsleepnet/l1_conv/conv1d/weights:0 (50, 1, 1, 64)
deepsleepnet/l1_conv/bn/beta:0 (64,)
deepsleepnet/l1_conv/bn/gamma:0 (64,)
deepsleepnet/l1_conv/bn/moving_mean:0 (64,)
deepsleepnet/l1_conv/bn/moving_variance:0 (64,)
deepsleepnet/l4_conv/conv1d/weights:0 (8, 1, 64, 128)
deepsleepnet/l4_conv/bn/beta:0 (128,)
deepsleepnet/l4_conv/bn/gamma:0 (128,)
deepsleepnet/l4_conv/bn/moving_mean:0 (128,)
deepsleepnet/l4_conv/bn/moving_variance:0 (128,)
deepsleepnet/l5_conv/conv1d/weights:0 (8, 1, 128, 128)
deepsleepnet/l5_conv/bn/beta:0 (128,)
deepsleepnet/l5_conv/bn/gamma:0 (128,)
deepsleepnet/l5_conv/bn/moving_mean:0 (128,)
deepsleepnet/l5_conv/bn/moving_variance:0 (128,)
deepsleepnet/l6_conv/conv1d/weights:0 (8, 1, 128, 128)
deepsleepnet/l6_conv/bn/beta:0 (128,)
deepsleepnet/l6_conv/bn/gamma:0 (128,)
deepsleepnet/l6_conv/bn/moving_mean:0 (128,)
deepsleepnet/l6_conv/bn/moving_variance:0 (128,)
deepsleepnet/l9_conv/conv1d/weights:0 (400, 1, 1, 64)
deepsleepnet/l9_conv/bn/beta:0 (64,)
deepsleepnet/l9_conv/bn/gamma:0 (64,)
deepsleepnet/l9_conv/bn/moving_mean:0 (64,)
deepsleepnet/l9_conv/bn/moving_variance:0 (64,)
deepsleepnet/l12_conv/conv1d/weights:0 (6, 1, 64, 128)
deepsleepnet/l12_conv/bn/beta:0 (128,)
deepsleepnet/l12_conv/bn/gamma:0 (128,)
deepsleepnet/l12_conv/bn/moving_mean:0 (128,)
deepsleepnet/l12_conv/bn/moving_variance:0 (128,)
deepsleepnet/l13_conv/conv1d/weights:0 (6, 1, 128, 128)
deepsleepnet/l13_conv/bn/beta:0 (128,)
deepsleepnet/l13_conv/bn/gamma:0 (128,)
deepsleepnet/l13_conv/bn/moving_mean:0 (128,)
deepsleepnet/l13_conv/bn/moving_variance:0 (128,)
deepsleepnet/l14_conv/conv1d/weights:0 (6, 1, 128, 128)
deepsleepnet/l14_conv/bn/beta:0 (128,)
deepsleepnet/l14_conv/bn/gamma:0 (128,)
deepsleepnet/l14_conv/bn/moving_mean:0 (128,)
deepsleepnet/l14_conv/bn/moving_variance:0 (128,)
deepsleepnet/l19_softmax_linear/weights:0 (3072, 5)
deepsleepnet/l19_softmax_linear/biases:0 (5,)
Pretrain Params
deepfeaturenet/l5_conv/conv1d/weights/Adam:0
deepfeaturenet/l6_conv/bn/beta/Adam:0
deepfeaturenet/l9_conv/bn/gamma/Adam:0
deepfeaturenet/l6_conv/bn/gamma/Adam:0
deepfeaturenet/l1_c```onv/conv1d/weights:0
deepfeaturenet/l5_conv/bn/beta/Adam_1:0
deepfeaturenet/l5_conv/bn/moving_mean:0
deepfeaturenet/l5_conv/conv1d/weights/Adam_1:0
deepfeaturenet/l6_conv/bn/moving_mean:0
deepfeaturenet/l4_conv/conv1d/weights/Adam:0
deepfeaturenet/l4_conv/bn/beta/Adam_1:0
deepfeaturenet/l4_conv/bn/gamma/Adam:0
deepfeaturenet/l19_softmax_linear/weights/Adam_1:0
deepfeaturenet/l13_conv/bn/gamma/Adam_1:0
deepfeaturenet/l4_conv/bn/beta:0
deepfeaturenet/l14_conv/bn/beta/Adam_1:0
deepfeaturenet/l9_conv/bn/moving_mean:0
deepfeaturenet/l14_conv/conv1d/weights:0
deepfeaturenet/l5_conv/bn/gamma:0
deepfeaturenet/l4_conv/bn/gamma/Adam_1:0
deepfeaturenet/l9_conv/bn/beta/Adam:0
deepfeaturenet/l5_conv/bn/beta/Adam:0
deepfeaturenet/l6_conv/conv1d/weights/Adam_1:0
deepfeaturenet/l4_conv/bn/gamma:0
deepfeaturenet/l12_conv/bn/gamma/Adam:0
deepfeaturenet/l12_conv/bn/beta/Adam_1:0
deepfeaturenet/l9_conv/bn/moving_variance:0
deepfeaturenet_2/global_step:0
deepfeaturenet/l13_conv/conv1d/weights/Adam_1:0
deepfeaturenet/l9_conv/bn/beta:0
deepfeaturenet/l12_conv/conv1d/weights/Adam:0
deepfeaturenet/l6_conv/conv1d/weights:0
deepfeaturenet/l4_conv/bn/beta/Adam:0
deepfeaturenet/l19_softmax_linear/biases/Adam_1:0
deepfeaturenet/l14_conv/conv1d/weights/Adam:0
deepfeaturenet/l4_conv/bn/moving_mean:0
deepfeaturenet/l6_conv/bn/beta/Adam_1:0
deepfeaturenet/l13_conv/bn/moving_mean:0
deepfeaturenet/l13_conv/bn/gamma:0
deepfeaturenet/l1_conv/bn/gamma:0
deepfeaturenet/l14_con```v/bn/gamma/Adam_1:0
deepfeaturenet/l9_conv/conv1d/weights:0
deepfeaturenet/l12_conv/conv1d/weights/Adam_1:0
deepfeaturenet/l1_conv/bn/gamma/Adam:0
deepfeaturenet/l9_conv/conv1d/weights/Adam_1:0
deepfeaturenet/l6_conv/bn/moving_variance:0
deepfeaturenet/l5_conv/bn/gamma/Adam:0
deepfeaturenet/l9_conv/conv1d/weights/Adam:0
deepfeaturenet/l19_softmax_linear/weights/Adam:0
deepfeaturenet/l6_conv/bn/gamma:0
deepfeaturenet/l12_conv/bn/beta/Adam:0
deepfeaturenet/l13_conv/bn/moving_variance:0
deepfeaturenet/l4_conv/conv1d/weights:0
deepfeaturenet/l13_conv/bn/gamma/Adam:0
deepfeaturenet/l19_softmax_linear/biases/Adam:0
deepfeaturenet/l13_conv/bn/beta/Adam_1:0
deepfeaturenet/l5_conv/bn/moving_variance:0
deepfeaturenet/l9_conv/bn/gamma:0
deepfeaturenet/l1_conv/bn/moving_mean:0
deepfeaturenet/l4_conv/bn/moving_variance:0
deepfeaturenet/l19_softmax_linear/weights:0
beta1_power:0
deepfeaturenet/l1_conv/bn/moving_variance:0
deepfeaturenet/l6_conv/conv1d/weights/Adam:0
deepfeaturenet/l13_conv/bn/beta:0
deepfeaturenet/l12_conv/bn/moving_mean:0
deepfeaturenet/l14_conv/bn/beta:0
deepfeaturenet/l14_conv/bn/gamma:0
deepfeaturenet/l14_conv/bn/moving_mean:0
deepfeaturenet/l13_conv/conv1d/weights:0
deepfeaturenet/l5_conv/conv1d/weights:0
deepfeaturenet/l12_conv/bn/beta:0
deepfeaturenet/l1_conv/bn/beta/Adam_1:0
deepfeaturenet/l4_conv/conv1d/weights/Adam_1:0
deepfeaturenet/l12_conv/conv1d/weights:0
beta2_power:0
deepfeaturenet/l1_conv/bn/beta:0
deepfeaturenet/l13_conv/bn/beta/Adam:0
deepfeaturenet/l14_conv/conv1d/weights/Adam_1:0
deepfeaturenet/l13_conv/conv1d/weights/Adam:0
deepfeaturenet/l6_conv/bn/gamma/Adam_1:0
deepfeaturenet/l6_conv/bn/beta:0
deepfeaturenet/l1_conv/conv1d/weights/Adam:0
deepfeaturenet/l1_conv/bn/beta/Adam:0
deepfeaturenet/l9_conv/bn/beta/Adam_1:0
deepfeaturenet/l12_conv/bn/gamma:0
deepfeaturenet/l1_conv/bn/gamma/Adam_1:0
deepfeaturenet/l14_conv/bn/moving_variance:0
deepfeaturenet/l9_conv/bn/gamma/Adam_1:0
deepfeaturenet/l12_conv/bn/moving_variance:0
deepfeaturenet/l5_conv/bn/gamma/Adam_1:0
deepfeaturenet/l12_conv/bn/gamma/Adam_1:0
deepfeaturenet/l5_conv/bn/beta:0
deepfeaturenet/l14_conv/bn/beta/Adam:0
deepfeaturenet/l1_conv/conv1d/weights/Adam_1:0
deepfeaturenet/l14_conv/bn/gamma/Adam:0
deepfeaturenet/l19_softmax_linear/biases:0
Pretrained parameters:
Optimizing parameters:
deepsleepnet/l9_conv/conv1d/weights:0
deepsleepnet/l5_conv/bn/gamma:0
deepsleepnet/l9_conv/bn/gamma:0
deepsleepnet/l5_conv/conv1d/weights:0
deepsleepnet/l6_conv/bn/beta:0
deepsleepnet/l4_conv/conv1d/weights:0
deepsleepnet/l13_conv/conv1d/weights:0
deepsleepnet/l6_conv/conv1d/weights:0
deepsleepnet/l4_conv/bn/gamma:0
deepsleepnet/l14_conv/conv1d/weights:0
deepsleepnet/l1_conv/bn/beta:0
deepsleepnet/l4_conv/bn/beta:0
deepsleepnet/l9_conv/bn/beta:0
deepsleepnet/l5_conv/bn/beta:0
deepsleepnet/l1_conv/conv1d/weights:0
deepsleepnet/l13_conv/bn/beta:0
deepsleepnet/l19_softmax_linear/weights:0
deepsleepnet/l12_conv/bn/gamma:0
deepsleepnet/l6_conv/bn/gamma:0
deepsleepnet/l1_conv/bn/gamma:0
deepsleepnet/l13_conv/bn/gamma:0
deepsleepnet/l12_conv/conv1d/weights:0
deepsleepnet/l12_conv/bn/beta:0
deepsleepnet/l19_softmax_linear/biases:0
deepsleepnet/l14_conv/bn/gamma:0
deepsleepnet/l14_conv/bn/beta:0
Trainable Variables:
deepsleepnet/l1_conv/conv1d/weights:0 (50, 1, 1, 64)
deepsleepnet/l1_conv/bn/beta:0 (64,)
deepsleepnet/l1_conv/bn/gamma:0 (64,)
deepsleepnet/l4_conv/conv1d/weights:0 (8, 1, 64, 128)
deepsleepnet/l4_conv/bn/beta:0 (128,)
deepsleepnet/l4_conv/bn/gamma:0 (128,)
deepsleepnet/l5_conv/conv1d/weights:0 (8, 1, 128, 128)
deepsleepnet/l5_conv/bn/beta:0 (128,)
deepsleepnet/l5_conv/bn/gamma:0 (128,)
deepsleepnet/l6_conv/conv1d/weights:0 (8, 1, 128, 128)
deepsleepnet/l6_conv/bn/beta:0 (128,)
deepsleepnet/l6_conv/bn/gamma:0 (128,)
deepsleepnet/l9_conv/conv1d/weights:0 (400, 1, 1, 64)
deepsleepnet/l9_conv/bn/beta:0 (64,)
deepsleepnet/l9_conv/bn/gamma:0 (64,)
deepsleepnet/l12_conv/conv1d/weights:0 (6, 1, 64, 128)
deepsleepnet/l12_conv/bn/beta:0 (128,)
deepsleepnet/l12_conv/bn/gamma:0 (128,)
deepsleepnet/l13_conv/conv1d/weights:0 (6, 1, 128, 128)
deepsleepnet/l13_conv/bn/beta:0 (128,)
deepsleepnet/l13_conv/bn/gamma:0 (128,)
deepsleepnet/l14_conv/conv1d/weights:0 (6, 1, 128, 128)
deepsleepnet/l14_conv/bn/beta:0 (128,)
deepsleepnet/l14_conv/bn/gamma:0 (128,)
deepsleepnet/l19_softmax_linear/weights:0 (3072, 5)
deepsleepnet/l19_softmax_linear/biases:0 (5,)
All Variables:
deepsleepnet/l1_conv/conv1d/weights:0 (50, 1, 1, 64)
deepsleepnet/l1_conv/bn/beta:0 (64,)
deepsleepnet/l1_conv/bn/gamma:0 (64,)
deepsleepnet/l1_conv/bn/moving_mean:0 (64,)
deepsleepnet/l1_conv/bn/moving_variance:0 (64,)
deepsleepnet/l4_conv/conv1d/weights:0 (8, 1, 64, 128)
deepsleepnet/l4_conv/bn/beta:0 (128,)
deepsleepnet/l4_conv/bn/gamma:0 (128,)
deepsleepnet/l4_conv/bn/moving_mean:0 (128,)
deepsleepnet/l4_conv/bn/moving_variance:0 (128,)
deepsleepnet/l5_conv/conv1d/weights:0 (8, 1, 128, 128)
deepsleepnet/l5_conv/bn/beta:0 (128,)
deepsleepnet/l5_conv/bn/gamma:0 (128,)
deepsleepnet/l5_conv/bn/moving_mean:0 (128,)
deepsleepnet/l5_conv/bn/moving_variance:0 (128,)
deepsleepnet/l6_conv/conv1d/weights:0 (8, 1, 128, 128)
deepsleepnet/l6_conv/bn/beta:0 (128,)
deepsleepnet/l6_conv/bn/gamma:0 (128,)
deepsleepnet/l6_conv/bn/moving_mean:0 (128,)
deepsleepnet/l6_conv/bn/moving_variance:0 (128,)
deepsleepnet/l9_conv/conv1d/weights:0 (400, 1, 1, 64)
deepsleepnet/l9_conv/bn/beta:0 (64,)
deepsleepnet/l9_conv/bn/gamma:0 (64,)
deepsleepnet/l9_conv/bn/moving_mean:0 (64,)
deepsleepnet/l9_conv/bn/moving_variance:0 (64,)
deepsleepnet/l12_conv/conv1d/weights:0 (6, 1, 64, 128)
deepsleepnet/l12_conv/bn/beta:0 (128,)
deepsleepnet/l12_conv/bn/gamma:0 (128,)
deepsleepnet/l12_conv/bn/moving_mean:0 (128,)
deepsleepnet/l12_conv/bn/moving_variance:0 (128,)
deepsleepnet/l13_conv/conv1d/weights:0 (6, 1, 128, 128)
deepsleepnet/l13_conv/bn/beta:0 (128,)
deepsleepnet/l13_conv/bn/gamma:0 (128,)
deepsleepnet/l13_conv/bn/moving_mean:0 (128,)
deepsleepnet/l13_conv/bn/moving_variance:0 (128,)
deepsleepnet/l14_conv/conv1d/weights:0 (6, 1, 128, 128)
deepsleepnet/l14_conv/bn/beta:0 (128,)
deepsleepnet/l14_conv/bn/gamma:0 (128,)
deepsleepnet/l14_conv/bn/moving_mean:0 (128,)
deepsleepnet/l14_conv/bn/moving_variance:0 (128,)
deepsleepnet/l19_softmax_linear/weights:0 (3072, 5)
deepsleepnet/l19_softmax_linear/biases:0 (5,)
trainable_variables
[<tf.Variable 'deepsleepnet/l1_conv/conv1d/weights:0' shape=(50, 1, 1, 64) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l1_conv/bn/beta:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l1_conv/bn/gamma:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l4_conv/conv1d/weights:0' shape=(8, 1, 64, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l4_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l4_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l5_conv/conv1d/weights:0' shape=(8, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l5_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l5_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l6_conv```/conv1d/weights:0' shape=(8, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l6_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l6_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l9_conv/conv1d/weights:0' shape=(400, 1, 1, 64) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l9_conv/bn/beta:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l9_conv/bn/gamma:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l12_conv/conv1d/weights:0' shape=(6, 1, 64, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l12_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l12_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l13_conv/conv1d/weights:0' shape=(6, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l13_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l13_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l14_conv/conv1d/weights:0' shape=(6, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l14_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l14_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l19_softmax_linear/weights:0' shape=(3072, 5) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l19_softmax_linear/biases:0' shape=(5,) dtype=float32_ref>]
train_vars1
[]
train_vars2
[<tf.Variable 'deepsleepnet/l9_conv/conv1d/weights:0' shape=(400, 1, 1, 64) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l5_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l9_conv/bn/gamma:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l5_conv/conv1d/weights:0' shape=(8, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l6_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l4_conv/conv1d/weights:0' shape=(8, 1, 64, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l13_conv/conv1d/weights:0' shape=(6, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l6_conv/conv1d/weights:0' shape=(8, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l4_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l14_conv/conv1d/weights:0' shape=(6, 1, 128, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l1_conv/bn/beta:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l4_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l9_conv/bn/beta:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l5_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l1_conv/conv1d/weights:0' shape=(50, 1, 1, 64) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l13_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l19_softmax_linear/weights:0' shape=(3072, 5) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l12_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l6_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l1_conv/bn/gamma:0' shape=(64,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l13_con```v/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l12_conv/conv1d/weights:0' shape=(6, 1, 64, 128) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l12_conv/bn/beta:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l19_softmax_linear/biases:0' shape=(5,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l14_conv/bn/gamma:0' shape=(128,) dtype=float32_ref>, <tf.Variable 'deepsleepnet/l14_conv/bn/beta:0' shape=(128,) dtype=float32_ref>]
Traceback (most recent call last):
File "train.py", line 95, in <module>
tf.app.run()
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/platform/app.py", line 124, in run
_sys.exit(main(argv))
File "train.py", line 90, in main
n_epochs=FLAGS.finetune_epochs
File "train.py", line 72, in finetune
resume=FLAGS.resume
File "/home/Abhay/Documents/SleepInterfacing/DeepSleepNet/deepsleepnet/deepsleep/trainer.py", line 648, in finetune
clip_value=10.0
File "/home/Abhay/Documents/SleepInterfacing/DeepSleepNet/deepsleepnet/deepsleep/optimize.py", line 60, in adam_clipping_list_lr
apply_gradient_op = opt.apply_gradients(gvs)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/training/optimizer.py", line 493, in apply_gradients
raise ValueError("No variables provided.")
ValueError: No variables provided.
Please go through this and let me know what the next steps of debugging are. Thanks!
Sorry for getting back to you late.
I think the problem should be with the name of the class or the name of the variable that you used as:
train_vars1
should contain the parameters of the "representation learning" part of the network, andtrain_vars2
should contain the parameters of the "sequence residual learning" part of the network.This would allow us to use different learning rate for each part of the whole model.
I still cannot find which part of your code cause this problem. I recommend you to try to run my code with the previous version of Tensorflow. You can create another Python virtual environment (e.g., using Miniconda2), and then install the required packages with the same version mentioned in the README. Then print out these train_vars1
and train_vars2
. I believe you would be able to identify where is the problem.
Hope this help.
Hi Akara, I successfully updated the full code of deepsleepnet to tensorflow 1.5. There happens to be an error in optimization after the completion of convolutional pre-training. To exactly point out where the error is, I printed the optimizing parameters as shown in the screenshot attached. The list train_vars2 happens to be empty which is why the opt.applygradients(gvs) is giving us ValueError: No variables provided. in the second run of that loop.
Here is the traceback.
Please help me out in resolving this error.
Best Regards, Abhay.