STT-tensorflow/tensorflow/compiler/tf2xla
A. Unique TensorFlower 411f0d2800 Fix ResolveInputDynamismIntoPredVector to consider values dynamic on failure to resolve dynamism
ResolveInputDynamismIntoPred correctly handles this.

Dynamism computation could fail if the input depends on some unsupported ops like custom-call. In such cases, all elements for the input should conservatively be marked as dynamic and not static.

PiperOrigin-RevId: 361161831
Change-Id: I7f6fcbd29fc1a2562af26d9682eaf1f02e11f6a7
2021-03-05 09:38:30 -08:00
..
cc Add target environment constraints to a subset of TensorFlow packages and targets. 2020-09-21 10:57:01 -07:00
g3doc
kernels Add MlirXlaOpKernel, which is used to implement XlaOpKernels using MLIR legalization. 2021-03-03 07:27:30 -08:00
lib Add target environment constraints to a subset of TensorFlow packages and targets. 2020-09-21 10:57:01 -07:00
ops Add a backwards compatibility note to the documentation for XLA TF ops 2021-03-02 01:46:41 -08:00
python Add a backwards compatibility note to the documentation for XLA TF ops 2021-03-02 01:46:41 -08:00
BUILD XLA value inference. 2021-03-03 14:17:48 -08:00
const_analysis_test.cc Rollback of rollback of disabling XLA:CPU/GPU devices by default. 2020-08-11 13:06:59 -07:00
const_analysis.cc [TF2XLA] [NFC] Remove the comment which was already done 2021-02-03 16:46:44 -08:00
const_analysis.h
cpu_function_runtime_test.cc
frontend_attributes_util.cc
frontend_attributes_util.h
functionalize_cond_test.cc Add capability to restrict functionalization via node filter 2020-06-29 10:56:59 -07:00
functionalize_cond.cc XLA Parallel reduce. 2020-08-27 15:21:33 -07:00
functionalize_cond.h Add capability to restrict functionalization via node filter 2020-06-29 10:56:59 -07:00
functionalize_control_flow_pass_registration.cc
functionalize_control_flow_test.cc Also functionalize control flow in functions for UpgradeLegacyGraph 2020-08-14 17:04:03 -07:00
functionalize_control_flow_util.cc Add capability to restrict functionalization via node filter 2020-06-29 10:56:59 -07:00
functionalize_control_flow_util.h Add capability to restrict functionalization via node filter 2020-06-29 10:56:59 -07:00
functionalize_control_flow.cc XLA Parallel reduce. 2020-08-27 15:21:33 -07:00
functionalize_control_flow.h Also functionalize control flow in functions for UpgradeLegacyGraph 2020-08-14 17:04:03 -07:00
functionalize_while.cc XLA Parallel reduce. 2020-08-27 15:21:33 -07:00
functionalize_while.h Add capability to restrict functionalization via node filter 2020-06-29 10:56:59 -07:00
fused_batchnorm_reserve_space_test.cc Rollback of rollback of disabling XLA:CPU/GPU devices by default. 2020-08-11 13:06:59 -07:00
graph_compiler_util.cc [-Wsign-compare] warning fixes batch 9 2020-07-07 01:37:54 +00:00
graph_compiler_util.h
graph_compiler.cc [TF2XLA] Support must-be-constant resource variables for compilation 2020-11-09 15:04:58 -08:00
graph_compiler.h
host_compute_metadata.proto
literal_util_test.cc
literal_util.cc tf2xla directory resolutions 2020-07-26 22:06:02 +00:00
literal_util.h
mlir_bridge_pass_registration.cc Add support of fallback functionality to MlirV1CompatGraphOptimizationPass. 2021-03-04 15:58:42 -08:00
mlir_bridge_pass.cc Lookup functions in optional FunctionLibraryDefinition when checking for supported features for MLIR bridge. 2021-03-05 09:34:14 -08:00
mlir_bridge_pass.h Lookup functions in optional FunctionLibraryDefinition when checking for supported features for MLIR bridge. 2021-03-05 09:34:14 -08:00
mlir_tf2xla.cc Propagate unknown rank of tensor shape information to MLIR 2020-11-17 00:38:39 -08:00
mlir_xla_op_kernel.cc Add MlirXlaOpKernel, which is used to implement XlaOpKernels using MLIR legalization. 2021-03-03 07:27:30 -08:00
mlir_xla_op_kernel.h Add MlirXlaOpKernel, which is used to implement XlaOpKernels using MLIR legalization. 2021-03-03 07:27:30 -08:00
rearrange_function_argument.cc [TF2XLA] Look for stack traces in global FunctionLibraryDefinition if they are not found in local FunctionLibraryDefinition 2021-01-26 19:51:52 -08:00
rearrange_function_argument.h [TF2XLA] Look for stack traces in global FunctionLibraryDefinition if they are not found in local FunctionLibraryDefinition 2021-01-26 19:51:52 -08:00
resource_operation_table_test.cc
resource_operation_table.cc Internal change 2020-09-17 21:37:45 -07:00
resource_operation_table.h
resource_util_test.cc
resource_util.cc
resource_util.h
shape_util.cc Reuse unknown sentinel when not static 2021-01-12 10:40:05 -08:00
shape_util.h
sharding_util_test.cc Update DistributedTPURewritePass to populate OpMetadata in OpSharding. 2021-01-19 10:26:20 -08:00
sharding_util.cc Update DistributedTPURewritePass to populate OpMetadata in OpSharding. 2021-01-19 10:26:20 -08:00
sharding_util.h Update DistributedTPURewritePass to populate OpMetadata in OpSharding. 2021-01-19 10:26:20 -08:00
side_effect_util.cc [XLA] Break dependencies of XLA on TF. 2021-01-19 14:07:56 -08:00
side_effect_util.h [XLA] Break dependencies of XLA on TF. 2021-01-19 14:07:56 -08:00
test_util.cc
test_util.h
tf2xla_supported_ops_main.cc
tf2xla_supported_ops.cc
tf2xla_supported_ops.h
tf2xla_test.cc
tf2xla_util_test.cc Update DistributedTPURewritePass to populate OpMetadata in OpSharding. 2021-01-19 10:26:20 -08:00
tf2xla_util.cc [TF2XLA] Propagate stack traces properly for more functions 2021-01-25 16:21:46 -08:00
tf2xla_util.h Restrict propagation of internal attributes in functionalization 2020-07-01 13:41:52 -07:00
tf2xla.cc Remove an obsolete TODO since allowed_device has been deprecated. 2020-08-20 15:54:51 -07:00
tf2xla.h
tf2xla.proto
type_util.cc
type_util.h
xla_argument.cc Add uninitialized resource args to unsupported features for MLIR bridge 2021-02-16 19:12:45 -08:00
xla_argument.h Use proper header for absl::optional 2021-02-17 22:38:26 -08:00
xla_compilation_device.cc Update DistributedTPURewritePass to populate OpMetadata in OpSharding. 2021-01-19 10:26:20 -08:00
xla_compilation_device.h
xla_compiled_cpu_function.cc
xla_compiled_cpu_function.h
xla_compiler_test.cc Update DistributedTPURewritePass and sharding util to populate OpMetadata in OpSharding. 2021-01-14 13:56:17 -08:00
xla_compiler.cc Lookup functions in optional FunctionLibraryDefinition when checking for supported features for MLIR bridge. 2021-03-05 09:34:14 -08:00
xla_compiler.h Fix enable MLIR bridge logic in xla_compiler 2020-11-30 15:25:39 -08:00
xla_context.cc [TF2XLA] Display Python location information in XLA metadata 2020-12-10 16:53:44 -08:00
xla_context.h [TF2XLA] Display Python location information in XLA metadata 2020-12-10 16:53:44 -08:00
xla_cpu_backend.cc
xla_expression_test.cc [TF2XLA] Support must-be-constant resource variables for compilation 2020-11-09 15:04:58 -08:00
xla_expression.cc XLA value inference. 2021-03-03 14:17:48 -08:00
xla_expression.h Propagate value bounds into control flow. 2021-02-17 15:42:19 -08:00
xla_gpu_backend.cc
xla_helpers.cc [XLA:SPMD] Minor fixes and utils for manual sharding 2020-12-01 13:37:36 -08:00
xla_helpers.h [TF2XLA] [NFC] Break apart the [TF2XLA/MLIR] -> xla_compiler dependency edge 2020-07-28 13:36:06 -07:00
xla_jit_compiled_cpu_function_test.cc
xla_jit_compiled_cpu_function.cc
xla_jit_compiled_cpu_function.h
xla_op_kernel.cc Fix ResolveInputDynamismIntoPredVector to consider values dynamic on failure to resolve dynamism 2021-03-05 09:38:30 -08:00
xla_op_kernel.h Mark XlaOpKernelContext::InputShape methods as deprecated 2021-02-25 19:11:07 -08:00
xla_op_registry_test.cc
xla_op_registry.cc Use TryGetNodeAttr instead of GetNodeAttr in a few places 2021-02-15 14:26:06 -08:00
xla_op_registry.h [TF2XLA] Remove the serialization of CustomKernelCreator, since there is only one, and we won't add new ones 2020-09-22 10:59:59 -07:00
xla_resource.cc [TF2XLA] Support must-be-constant resource variables for compilation 2020-11-09 15:04:58 -08:00
xla_resource.h [TF2XLA] Support must-be-constant resource variables for compilation 2020-11-09 15:04:58 -08:00
xla_tpu_backend.cc [TPU] Colocate tpu_compilation_device registration with other registration in compiler/tf2xla 2021-01-22 16:51:35 -08:00