pytorch / xla

3 Pull requests merged by 3 people

• Enable tpuvm pod support

  #2782 merged Feb 27, 2021

• 1VM: Unbreak libtpu.so API, XLA, profiler

  #2794 merged Feb 27, 2021

• Add version selection snippet for r1.8

  #2788 merged Feb 25, 2021

2 Pull requests opened by 2 people

• [r1.8] Cherrypick PR2794

  #2797 opened Feb 27, 2021

• [r.18] Backporting "Enable tpuvm pod support (#2782)"

  #2798 opened Feb 27, 2021

5 Issues closed by 5 people

• Fixing dataset and model loading 8 times in transformers library

  #2792 closed Feb 26, 2021

• Parameter server

  #2793 closed Feb 25, 2021

• non_blocking in .to() of XLA tensor

  #2791 closed Feb 22, 2021

• Model training works on GPU but not on Pytorch/XLA

  #2785 closed Feb 21, 2021

• NormFreeNets are not working in pytorch XLA

  #2776 closed Feb 20, 2021

4 Issues opened by 3 people

• Hang when no rendezvous / optimizer_step is used.

  #2796 opened Feb 26, 2021

• Unable to train RNN models with BucketIterator using TPUs

  #2795 opened Feb 26, 2021

• please lower std_mean and var_mean

  #2790 opened Feb 20, 2021

• Exception in device=TPU:3: Input contains NaN, infinity or a value too large for dtype('float32').

  #2789 opened Feb 20, 2021

2 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• Multiprocessing with GPUs setup for workers has errors

  #2758 commented on Feb 22, 2021 • 2 new comments

• Sync the tensors before an .item() call.

  #2770 commented on Feb 23, 2021 • 1 new comment