Kurmanji (Kurdish) trained model for Stanza (UD 2.9)

# Custom model for Kurmanji (Kurdish)
Trained on 2.9, using kmr_mg and word vec from fasttext. train and test are the same data !!!

Preparing data for UD_Kurmanji-MG: kmr_mg, kmr
Train/dev split not present.  Randomly splitting train file
754 total sentences found: 604 in train, 150 in dev.
14 unique MWTs found in data
7 unique MWTs found in data
14 unique MWTs found in data

## Score

TOKENIZER
UD_Kurmanji-MG
98.75 100.00 98.64

MWT
UD_Kurmanji-MG
100.00

Lemma score:
kmr_mg 95.09

POS score
2022-05-13 13:00:24 INFO: Start evaluation...
2022-05-13 13:00:24 INFO: UPOS    XPOS    UFeats    AllTags
2022-05-13 13:00:24 INFO: 94.19    93.95    86.30    85.36
2022-05-13 13:00:24 INFO: Tagger score:
2022-05-13 13:00:24 INFO: kmr_mg 85.36
2022-05-13 13:00:24 INFO: Finished running dev set on
UD_Kurmanji-MG
94.19 93.95 86.30 85.36

Dep score
2022-05-13 15:02:42 INFO: Start evaluation...
2022-05-13 15:02:43 INFO: LAS    MLAS    BLEX
2022-05-13 15:02:43 INFO: 76.38    67.48    72.02
2022-05-13 15:02:43 INFO: Parser score:
2022-05-13 15:02:43 INFO: kmr_mg 76.38
2022-05-13 15:02:43 INFO: Finished running dev set on
UD_Kurmanji-MG
83.28 76.38 72.02 67.48 72.02