Reed

@kaixih let me know if you need more information than the error message above. If so I can work on getting an example to reproduce the error.

This is failing internally. The issue is the `cublas_lt_header` build rule does not have build dependencies for all the headers it uses. This doesn't fail externally so unfortunately you cannot...

@kaixih Unfortunately the TensorFlow presubmits disallows any protobuf files to be included in a kernel header file, but you included `tensorflow/core/protobuf/autotuning.pb.h` in the file `tensorflow/core/kernels/conv_ops_gpu.h`. I'm not entierly sure what...

Also, in `MatmulParameters` within `tensorflow/core/util/autotune_maps/conv_parameters.h`, you added a new `LINT.IfChange(conv_parameters_version)` line. But there already is a different `LINT.IfChange(conv_parameters_version)` line earlier in the file. Can you fix, or just remove the...

Why reopen this? This has been merged with https://github.com/tensorflow/tensorflow/commit/23dd6bbe697393b97407829ecc32d88821484406.

Hmm I wonder why it wasn't marked as merged. I think the issue might have been I started manually merging this before you forced push a new commit. So I...

This is failing an internal test with ``` error: non-constant-expression cannot be narrowed from type 'dnn::ActivationMode' to 'double' in initializer list [-Wc++11-narrowing] ``` I can get more context if you...

Hmm looks like this issue has to be fixed in internal code. I'll try to fix and merge.

I beleive the issue is that in [preprocessing.py](https://github.com/tensorflow/benchmarks/blob/841fa31263364795dd1c0a6abf35029116765d4c/scripts/tf_cnn_benchmarks/preprocessing.py#L647), we have a `tf.constant` node with all the images, which takes up more than a gigabyte of space in the graph.pbtxt that's...

You're right, the `events.out.tfevents` is also huge. I'm not sure whether the graph definition is written to it. The issue goes away if I omit `--data_dir`. Also, in TensorBoard, I...