PiperOrigin-RevId: 262864322

PiperOrigin-RevId: 262861510

failures.

PiperOrigin-RevId: 262857756

PiperOrigin-RevId: 262855058

This is a part of a larger migration effort for tensorflow::tstring.
See: https://github.com/tensorflow/community/pull/91
PiperOrigin-RevId: 262850088

The pattern list is not modified by any of these APIs and should thus be passed with const.

PiperOrigin-RevId: 262844002

1/8/16 bit attrs.  NFC

PiperOrigin-RevId: 262843016

Add tf_device dialect that models TensorFlow's actions of launching computations on accelerator devices.

PiperOrigin-RevId: 262833506

PiperOrigin-RevId: 262828591

PiperOrigin-RevId: 262789479

PiperOrigin-RevId: 262787528

PiperOrigin-RevId: 262787524

The function is very specific, and has limited use.

PiperOrigin-RevId: 262782380

The existing implementation only works when the FullyConnected has none bias.
This CL is to fuse the Add to the FullyConnected when there is a bias operand.

PiperOrigin-RevId: 262779181

PiperOrigin-RevId: 262757814

The two patterns are actually to fold a sequence of Conv2D, BiasAdd/AddV2 and
Mul.  Technically, this sequence should be captured by a single pattern, but
the current TableGen tool couldn't infer the output type of the Conv2D op,
since its "depth" is 1. For this reason, a predicate to enable this pattern
only if the input is defined by a Conv2D.

PiperOrigin-RevId: 262757621

The current implementation only returns one element for the splat case, which often comes as a surprise; leading to subtle/confusing bugs. The new behavior will include an iterate over the full range of elements, as defined by the shaped type, by providing the splat value for each iterator index.

PiperOrigin-RevId: 262756780

Replace const ref of vectors with llvm::ArrayRef and pointers of TF executor ops with values in arguments of functions. Switch from passing around an OpBuilder to creating one locally when creating IslandOps and YieldOps.

PiperOrigin-RevId: 262756302

Updated char* tstring::data() to conform to C++11.

This is a part of a larger migration effort for tensorflow::tstring.
See: https://github.com/tensorflow/community/pull/91
PiperOrigin-RevId: 262752692

This aligns with the namespace already used for the dialect.

PiperOrigin-RevId: 262751499

PiperOrigin-RevId: 262751066

PiperOrigin-RevId: 262747813

PiperOrigin-RevId: 262742602

Ensure we can propagate error properly when setting "TF_ENABLE_EAGER_CLIENT_STREAMING_ENQUEUE=false".

PiperOrigin-RevId: 262737847

This is a part of a larger migration effort for tensorflow::tstring.
See: https://github.com/tensorflow/community/pull/91
PiperOrigin-RevId: 262737181

Move annotation's thread_local storage to .cc file and clean up dependencies to avoid a duplicate symbol error.

PiperOrigin-RevId: 262733487

PiperOrigin-RevId: 262726250

PiperOrigin-RevId: 262704925

PiperOrigin-RevId: 262704920

PiperOrigin-RevId: 262700718

PiperOrigin-RevId: 262696484

PiperOrigin-RevId: 262682685

There are currently several different terms used to refer to a parent IR unit in 'get' methods: getParent/getEnclosing/getContaining. This cl standardizes all of these methods to use 'getParent*'.

PiperOrigin-RevId: 262680287

PiperOrigin-RevId: 262676072

About 15x speedup.

PiperOrigin-RevId: 262675867

In declarative rewrite rules, a symbol can be bound to op arguments or
results in the source pattern, and it can be bound to op results in the
result pattern. This means given a symbol in the pattern, it can stands
for different things: op operand, op attribute, single op result,
op result pack. We need a better way to model this complexity so that
we can handle according to the specific kind a symbol corresponds to.

Created SymbolInfo class for maintaining the information regarding a
symbol. Also created a companion SymbolInfoMap class for a map of
such symbols, providing insertion and querying depending on use cases.

PiperOrigin-RevId: 262675515

PiperOrigin-RevId: 262675086

PiperOrigin-RevId: 262671025

PiperOrigin-RevId: 262668255

By default, all functions are exported including main. This flag will be used to support new use cases like conversion of speech models to tflite where we want to run specific inference graphs through the converter and export specific functions (e.g. Joint, Encoder etc.) to a separate flatbuffer.

PiperOrigin-RevId: 262667934