This is a follow up on #45917 - Previously we fixed test_serialization by

• avoiding the usage of filename in torch.save/load;
• kept one specific test to validate both file descriptor and filename code path for torch.save/load

However per @malfet 's comment we can also abstract out temp file creation step to support the

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

Like xarray, pandas supports attaching arbitrary metadata to DataFrames and persisting it across operations. https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.attrs.html

Dask could pretty easily implement this as well. We'd have a

• _Frame.attrs property. This would likely return self._meta.attrs.
• We'd verify that dd.from_pandas(data) correctly extracts attrs from `da

See: numba/numba#6368 (comment)

The values tested will be random for each invocation of the tests, because there is no RNG seeding. The RNG should be seeded for each test, so that values are stable.

Well, Gumbel Distribution is magical. Basically, given a sequence of K logits, i.e., "\log a_1, \log a_2, ..., \log a_K" and K independent gumbel random variables, i.e., "g_1, g_2, ..., g_K". We have

\argmax_i \log a_i + g_i ~ Categorical({a_i / sum(a)})

This gives you a very simple way to sampl

Support Series.median()

What happened:
xr.DataArray([1], coords=[('onecoord', [2])]).sel(onecoord=2).to_dataframe(name='name') raise an exception ValueError: no valid index for a 0-dimensional object

What you expected to happen:

the same behavior as: xr.DataArray([1], coords=[('onecoord', [2])]).to_dataframe(name='name')

Anything else we need to know?:

I see that the array after the select

hi,
if possible, please add these indicators as well:
TDI (Traders Dynamic Index)
chandelier exit
pivot points
BOP (balance of power)
CTM (Chande trend meter)
Coppock Curve
Correlation Coefficient
PMO (DecisionPoint Price Momentum Oscillator)
Ulcer Index
most of them except TDI are available on stockcharts.com
thanks