Swen Gross

For PB Models it _seems_ like you can resolve this by changing `self.listener.runner.model.save(self.args.model)` to `self.listener.runner.model.save(self.args.model+'/')` (train_incremental.py) training models with ./precise-train [params] modelname path/to/modelname-files resp. (re)training against random files with ./precise-train-incremental...

Regarding the PR ([from](https://github.com/andreselizondo-adestech/mycroft-precise); tf 2.2): At least my ArchLinux Virtualbox (x86_64;py3.8 (v)env) refuses to cooperate (same reason) **EDIT: this behaviour changed with editing setup.py** ``` install_requires=[ 'numpy', 'tensorflow-gpu==2.2.0rc', ```...

Let's chime in. As is you'd need to pass an english singular to get the localized plural. I guess there are some good cross-language arguments. Lets say you want to...

not to say that 04/11/2022 isn't a monday. edit: shoot, looked at may (__yet the code wouldn't care if day mismatches datum__) this one year offset also occurs at _de,...

Ok, got it. (one year offset) Since ``` temp = datetime.strptime(datestr, "%B %d") extractedDate = extractedDate.replace(hour=0, minute=0, second=0) ``` eg (if called on the same day) ``` >>>2021-05-01 00:00:00+02:00 (temp)...

dayOffset: The problem is that weekdays in on itself are treated as relative words (what is reflected in dayOffset from now) -or in other words durations. And those durations collide...

@ChanceNCounter Hell yeah! There's a lot to identify. And tackles (from what i can tell) those pitfalls i measly outlined. Don't sleep on that one. And you've said there's nothing...

I don't think this would be caught on any language since the tests always give an anchorDate ``` date = datetime(2017, 6, 27, 0, 0) [extractedDate, leftover] = extract_datetime(text, date)...

~~Hm.. probably a bad idea to restrict year to >2000.~~ ~~will replace it with >60 since this this would exclude minute (and ofc hour) time data~~

If this is a bug across the spectrum i will add specific lines to all parsers