Continue training on model

[vlangloisx3]

I've been having some issues with the --continueParams option. The code that I am using to continue training a model is below:

uuparser --outdir ewt_output --datadir ud-treebanks-v2.6 --include "en_ewt" --continue --continueParams=params.pickle

My assumption is that the params file is in the same folder as the outdir, but I get the error: No such file or directory: 'params.pickle'. I've also tried inserting the actual path, but get the error: UnicodeDecodeError: 'utf-8' codec can't decode byte 0x80 in position 0: invalid start byte.

Since neither of these seem to work, where does the Params file need to be in order to load it?

Thanks!
Val

[LoicGrobol]

Hi ! Can you paste the command line you used when using the actual path and the traceback for the UnicodeDecodeError?

[vlangloisx3]

Sure! Here is the command line I used when using the path:
uuparser --outdir ewt_output --datadir ud-treebanks-v2.6 --include "en_ewt" --continue --continueParams=/Users/vlanglois/Desktop/project/ewt_output/en_ewt/params.pickle

and here is the traceback:
Traceback (most recent call last): File "/Library/anaconda3/bin/uuparser", line 8, in <module> sys.exit(main()) File "/Library/anaconda3/lib/python3.7/site-packages/uuparser/parser.py", line 279, in main run(experiment,options) File "/Library/anaconda3/lib/python3.7/site-packages/uuparser/parser.py", line 36, in run stored_vocab, stored_options = pickle.load(paramsfp) File "/Library/anaconda3/lib/python3.7/codecs.py", line 322, in decode (result, consumed) = self._buffer_decode(data, self.errors, final) UnicodeDecodeError: 'utf-8' codec can't decode byte 0x80 in position 0: invalid start byte

[LoicGrobol]

(Sorry for the latency, I am really low-bandwidth on uuparser right now but I'll try to fix this :-)

Can you try installing from the master branch (using pip install git+https://github.com/UppsalaNLP/uuparser)? There were some changes since the last release, which I suspect won't make a difference for this but it wouldn't hurt to be sure and at least we'd have up-to-date line numbers this way

Actually, I think I found the issue, can you try installing from the fixup branch using pip install git+https://github.com/UppsalaNLP/uuparser@fix-25 and see if it fixes the issue?

[vlangloisx3]

I'm still getting the same error even after installing from the fixup branch. It could be something wrong on my end though.

(Thanks for taking the time to look into it!)

[LoicGrobol]

Oh, if you installed with the exact command I posted in an env where uuparser is already installed, it might not have been updated because I forgot the -U flag, could you retry with pip install -U git+https://github.com/UppsalaNLP/uuparser@fix-25? Sorry for the inconvenience

[vlangloisx3]

Yes that solved the issue! Thank you so much!

[LoicGrobol]

Great, I'll merge the fix and push a release before the end of the week, then. Thank you for catching this one :-)

[LoicGrobol]

Closed by #26