Lorenzo Stella

I like this proposal because: 1. One doesn't need additional imports. 2. The evaluation is unambiguously done on the predictor object, and not on a predictor OR an estimator (in...

@jaheba should we punt this to the 0.6 milestone?

I think this may apply to more than just Arrow/Parquet, but also JSONLines maybe. Also, there may be multiple columns that contain e.g. features that one would want to stack...

@fernandocamargoti are you using mxnet 1.6.0? And the `DeepAREstimator`? It’s very weird indeed that this shows up with `hybridize=True` only. One place where things could go wrong is this https://github.com/awslabs/gluon-ts/blob/0d963f7dc55ef866d86e33633a28d57dfab33adb/src/gluonts/distribution/neg_binomial.py#L114...

> TARGET: Count values, with a lot of zeros What’s the fraction of zero values, roughly? It would be nice if we could to reproduce the issue

> So, each time-series may start on a different date, but I've normalized all of them to finish on the same date. @fernandocamargoti as a side note: that that's not...

@fernandocamargoti also a minor note: the logs you pasted for hybridized=True suggest that you're using num_batches_per_epoch=4000, but you mentioned 10_000 (still, it seems to converge too slow even assuming that...

Oh I see, I hadn’t scrolled all the way to the right in the first log, sorry. I also agree on the zero padding you’re doing, since that’s a behavior...

@fernandocamargoti thanks for looking into it, this is really cool (but also not cool that the bug is there). Did you observe that with 1.4 and 1.5 both NaN didn’t...

@fernandocamargoti the issue you’re facing with mxnet 1.5 is a known one on Linux: https://github.com/apache/incubator-mxnet/issues/16135 This was fixed in 1.6 but the fix was not backported to 1.5 I believe....