🐛 Bug

Tensor averaging not implemented for complex types

To Reproduce

>>> import torch
>>> x = torch.ones(4, dtype=torch.complex128).cuda()
>>> torch.mean(x)
---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-108-34af2dc3edea> in <module>
      1 impo

Improve navigation of the I/O how-to by prepending a decision tree/flowchart using an image or graphviz, as suggested by @mattip.

Quick scanning, the following should be corrected:

• "Does Dask have a SQL interface": yes it now does (if perhaps not fully featured)
• paid support: add Coiled? others (saturn)?
• main page Supported By: there was discussion on how to add logos here, but I would say Coiled, Saturn, CapitalOne, Prefect should be here. Also, an older page Funding should be the same with more details, or just

PR #6447 adds a public API to get the maximum number of registers per thread (numba.cuda.Dispatcher.get_regs_per_thread()). There are other attributes that might be nice to provide - shared memory per block, local memory per thread, const memory usage, maximum block size.

These are all available in the FuncAttr named tuple: https://github.com/numba/numba/blob/master/numba/cuda/cudadrv/drive

Is there a way to train a bidirectional RNN (like LSTM or GRU) on trax nowadays?

The current version of the HANS dataset is missing the additional information provided for each example, including the sentence parses, heuristic and subcase.

Description

Move the adamw optimizer defined here to MXNet. https://github.com/dmlc/gluon-nlp/blob/20321598a732d254d4a9e036eecbe0ae8abb48a0/src/gluonnlp/optimizer.py#L31-L32

Support Series.median()

MCVE Code Sample

# Your code here
import numpy as np
import xarray as xr

data = np.zeros((10, 4))
example_xr = xr.DataArray(data, coords=[range(10), ["