Type AdamParameters
Namespace tensorflow.tpu.experimental
Parent _OptimizationParameters
Interfaces IAdamParameters
Optimization parameters for Adam with TPU embeddings. Pass this to
tf.estimator.tpu.experimental.EmbeddingConfigSpec
via the
`optimization_parameters` argument to set the optimizer and its parameters.
See the documentation for tf.estimator.tpu.experimental.EmbeddingConfigSpec
for more details. ```
estimator = tf.estimator.tpu.TPUEstimator(
...
embedding_config_spec=tf.estimator.tpu.experimental.EmbeddingConfigSpec(
...
optimization_parameters=tf.tpu.experimental.AdamParameters(0.1),
...))
```
Methods
Properties
Public static methods
AdamParameters NewDyn(object learning_rate, ImplicitContainer<T> beta1, ImplicitContainer<T> beta2, ImplicitContainer<T> epsilon, ImplicitContainer<T> lazy_adam, ImplicitContainer<T> sum_inside_sqrt, ImplicitContainer<T> use_gradient_accumulation, object clip_weight_min, object clip_weight_max)
Optimization parameters for Adam.
Parameters
-
object
learning_rate - a floating point value. The learning rate.
-
ImplicitContainer<T>
beta1 - A float value. The exponential decay rate for the 1st moment estimates.
-
ImplicitContainer<T>
beta2 - A float value. The exponential decay rate for the 2nd moment estimates.
-
ImplicitContainer<T>
epsilon - A small constant for numerical stability.
-
ImplicitContainer<T>
lazy_adam - Use lazy Adam instead of Adam. Lazy Adam trains faster. Please see `optimization_parameters.proto` for details.
-
ImplicitContainer<T>
sum_inside_sqrt - This improves training speed. Please see `optimization_parameters.proto` for details.
-
ImplicitContainer<T>
use_gradient_accumulation - setting this to `False` makes embedding gradients calculation less accurate but faster. Please see `optimization_parameters.proto` for details. for details.
-
object
clip_weight_min - the minimum value to clip by; None means -infinity.
-
object
clip_weight_max - the maximum value to clip by; None means +infinity.