experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Mark variables initialized in XLA, so Tensorflow also knows they're initialized.

Browse Source 
PiperOrigin-RevId: 276530886
Change-Id: I5c0d82dce44c81139efc766807fba1039f085de3
This commit is contained in:
Tres Popp 2019-10-24 11:31:54 -07:00 committed by TensorFlower Gardener
parent 95a6f4571c
commit f41d1939cf
1 changed files with 3 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
tensorflow/compiler/jit/xla_launch_util.cc
View File

@ -32,6 +32,7 @@ limitations under the License.
#include "tensorflow/core/framework/node_def_util.h"
#include "tensorflow/core/framework/op.h"
#include "tensorflow/core/framework/op_kernel.h"
#include "tensorflow/core/framework/resource_mgr.h"
#include "tensorflow/core/framework/tensor.h"
#include "tensorflow/core/framework/types.h"
#include "tensorflow/core/lib/core/errors.h"
@ -367,6 +368,7 @@ static Status SetBufferForResourceVarTensorUnderAllocateXlaTensors(
}
}
*variable_infos[i].var()->tensor() = output_tensor;
variable_infos[i].var()->is_initialized |= write.modified;
return Status::OK();
}
@ -540,6 +542,7 @@ Status XlaComputationLaunchContext::PopulateOutputs(
kernel->input_mapping, resource_var_snapshots, write.type,
write.shape, buffer, allocator);
*variable_infos[i].var()->tensor() = output_tensor;
variable_infos[i].var()->is_initialized |= write.modified;
}
++output_num;
}