ValueError: learning_rate type (<class 'float'>) not supported. learning_rate must be a training schedule (output of learning_rate_schedule() function) #1058
# Instantiate the trainer object to drive the model training
learning_rate = 0.02
learner = sgd(z.parameters, lr=learning_rate)
trainer = Trainer(z, loss, eval_error, [learner])
I'm getting error:
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-11-78dcd8830e76> in <module>()
1 # Instantiate the trainer object to drive the model training
2 learning_rate = 0.02
----> 3 learner = sgd(z.parameters, lr=learning_rate)
4 trainer = Trainer(z, loss, eval_error, [learner])
C:\src\cntk\bindings\python\cntk\utils\swig_helper.py in wrapper(*args, **kwds)
54 @wraps(f)
55 def wrapper(*args, **kwds):
---> 56 result = f(*args, **kwds)
57 map_if_possible(result)
58 return result
C:\src\cntk\bindings\python\cntk\learner.py in sgd(parameters, lr, l1_regularization_weight, l2_regularization_weight, gaussian_noise_injection_std_dev, gradient_clipping_threshold_per_sample, gradient_clipping_with_truncation)
330 Networks: Tricks of the Trade: Springer, 2012.
331 '''
--> 332 _verify_learning_rate_type(lr)
333 gaussian_noise_injection_std_dev = \
334 training_parameter_schedule(gaussian_noise_injection_std_dev, UnitType.minibatch)
C:\src\cntk\bindings\python\cntk\learner.py in _verify_learning_rate_type(learning_rate)
62 'learning_rate must be a training schedule '
63 '(output of learning_rate_schedule() function)'
---> 64 % type(learning_rate))
65
66 # an internal method to verify that the mometum schedule
ValueError: learning_rate type (<class 'float'>) not supported. learning_rate must be a training schedule (output of learning_rate_schedule() function)
Hi!
I'm running this example: https://github.com/Microsoft/CNTK/blob/master/Tutorials/CNTK_101_LogisticRegression.ipynb On Windows 10, 64 bit, on CPU. I built CNTK from source exactly like the way described in Wiki.
When running the example: