I'm trying to use adadelta optimizer in my training process but it crashed. It came to that I used different types of value in grad parameters.
System information
- OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Ubuntu 20.04
- TensorFlow installed from (source or binary): binary
- TensorFlow version (use command below): 2.8.0
- Python version: 3.7.11
Describe the current behavior
Crashed with error info.
Describe the expected behavior
Error info could show the wrong type and would not crash.
Standalone code to reproduce the issue
from tensorflow.python.eager import context
from tensorflow.python.framework import constant_op
from tensorflow.python.framework import dtypes
from tensorflow.python.ops import variables
from tensorflow.python.training import adadelta
if __name__ == "__main__":
num_updates = 4
for grad in [0.2, 0.1, 0.01]:
for lr in [1.0, 0.5, 0.1]:
var0_init = [1.0, 2.0]
var1_init = [3.0, 4.0]
var0 = variables.Variable(var0_init, dtype=dtypes.float32)
var1 = variables.Variable(var1_init, dtype=dtypes.float32)
grads = constant_op.constant([grad, grad], dtype=dtypes.float16) # will pass if use dtypes.float32
accum = 0.0
accum_update = 0.0
rho = 0.95
epsilon = 1e-08
adadelta_opt = adadelta.AdadeltaOptimizer(learning_rate=lr, rho=rho, epsilon=epsilon)
if (not context.executing_eagerly()):
adadelta_update = adadelta_opt.apply_gradients(zip([grads, grads], [var0, var1]))
slot = ([None] * 2)
slot_update = ([None] * 2)
for step in range(num_updates):
adadelta_opt.apply_gradients(zip([grads, grads], [var0, var1]))
Other info / logs Include any logs or source code that would be helpful to
2022-03-19 05:46:32.562381: F tensorflow/core/framework/tensor.cc:718] Check failed: dtype() == expected_dtype (1 vs. 19) half expected, got float
Aborted (core dumped)
I'm trying to use adadelta optimizer in my training process but it crashed. It came to that I used different types of value in grad parameters.
System information
Describe the current behavior
Crashed with error info.
Describe the expected behavior
Error info could show the wrong type and would not crash.
Standalone code to reproduce the issue
Other info / logs Include any logs or source code that would be helpful to