WISE MEMORY OPTIMIZER SERIAL NUMBERS UPGRADE

Upgrade Flatbuffers v2.0.5 from v1.12.0.

WISE MEMORY OPTIMIZER SERIAL NUMBERS UPDATE

tfl.scatter_nd now supports I1 for update arg.tf.unsortedsegmentprod op is supported.

tf.einsum is supported with multiple unknown shapes.

tensorflow::StatusOr will also become in the future alias to absl::StatusOr, so use StatusOr::value instead of StatusOr::ConsumeValueOrDie.

static_cast(error::Code::INVALID_ARGUMENT) or static_cast(code) for comparisons.

Use the free functions such as tensorflow::errors::IsInvalidArgument if needed.

Use the constructors such as tensorflow::errors:InvalidArgument to create status using an error code without accessing it.

Accessing tensorflow::error::Code fields is fine.

One MUST NOT access tensorflow::errors::Code fields.

Stop constructing Status objects from tensorflow::error::Code.

Use tensorflow::OkStatus() instead of tensorflow::Status::OK().

The C++ tensorflow::Code and tensorflow::Status will become aliases of respectively absl::StatusCode and absl::Status in some future release.

This is because it would produce the same values each time, which may not be intended.

An unseeded initializer will raise a warning if it is reused (called) multiple times.

For unseeded initializers ( seed=None), a random seed will be created and assigned at initializer creation (different initializer instances get different seeds).

Both seeded and unseeded initializers will always generate the same values every time they are called (for a given variable shape).

Keras initializers will now use stateless random ops to generate random numbers.

RNG behavior change for tf.keras.initializers.

If you decide to keep using the old optimizer, please explicitly change your optimizer to tf. Most users won't be affected by this change, but please check the API doc if any API used in your workflow is changed or deprecated, and make adaptions. The current tf. will continue to be supported as tf., e.g., tf.

tf. will graduate in Release 2.11, which means tf. will be an alias of tf.

Please update your imports accordingly, the old files will be removed in Release 2.11.

Some files in tensorflow/python/training have been moved to tensorflow/python/tracking and tensorflow/python/checkpoint.

Causal attention in and is now specified in the call() method via the use_causal_mask argument (rather than in the constructor), for consistency with other layers.

Wise memory optimizer serial numbers