System information

• Have I written custom code (as opposed to using a stock example script provided in TensorFlow): no.
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): osx 10.11.6
• TensorFlow installed from (source or binary): binary
• TensorFlow version (use command below): 1.8
• Python version: NA
• Bazel version (if compiling from source): NA
• **

The default value of said parameter is None. As I understand it, this will result in the Lambda layer's discarding of the previous mask. From the source code of the Lambda layer, the mask parameter may or may not be a callable. In the latter case, it takes parameters mask and inputs, but I do not understand how such a callable would behave or what it would yield. It would be great to have at

Computing actual WER values (using all samples) during training is too time consuming and not practical. An easy and more practical approach would be to calculate an WER estimate on base of a small random sub-set of the samples.

Hi, i testing face-api.js and trying loading a local image for make face similarity and is throw me "Blob is not defined" in my nodeJS application, here the function where i try computing descriptors:

face.env.monkeyPatch({ fetch, Blob });

const threshold = 0.6;
var descriptors = { desc1: null, desc2: null };

function computarDescriptor(aDesc, aUri) {
  // ~statics/pics/001_tmp.jpg

All the examples seem to implement data parallelism. Can I train a model that is too big for my card?

Keep getting this error when deploying kubeflow. I have made id anonymous by replacing my project ID with these words.

kubeflow: {"ResourceType":"container.v1.cluster","ResourceErrorCode":"400","ResourceErrorMessage":{"code":400,"message":"Project "PROJECT_ID" has no network named "default".","status":"INVALID_ARGUMENT","statusMessage":"Bad Request","requestPath":"https://container.googleap

ONNX needs an awesome list. All that exists now is a short list of tools, but the real list is already a bit longer.

MXNet already has one (with some ONNX stuff on it).