Bairen Yi

I don't think sharing 1 GPU between multiple applications is generally a good idea. Given TF takes control of the device global memory allocation, and all other resources that requires...

Since pthread_once fires function call twice. it is reasonable for us to file the issue twice. (Credit: https://www.zhihu.com/people/zhu-xiao-e/)

I wasn't looking into this for long time. You're free to use my code as it is.

This PR should be raised in core PyTorch as `make_fx` is moved there.

A little bit more context: this issue hits me when I try to trace through batch norm in training mode, as the number of tracked batches is incremented twice during...

In HCA mode, a veth is created separately from the IB device (using Calico or Flannel), and the IF binds to the netdev, e.g. eth0, is bridged to the host...

> @byronyi If we are going to contribute to addon first, do we need a RFC here? I guess the design was originally targeted to TF core. As @alextp said,...

> I think TensorFlow can provide a way to extend optimizers so that you can extend existing optimizers to handle your sparse weights. cc @omalleyt12 who proposes the new customizable...

@rz001 Hi, author of GDR here. Sorry that I just saw this issue. Do you still run into this error now?

> Hello, do you have any information about when optimizers will be allowed to perform global gradient clipping across several GPUs? Some of my work quite critically depends on this....