model improved

README.md

@@ -18,7 +18,7 @@ widget:
 
 ## Model Description
 
-This is a DeBERTa(V2) model pretrained on Japanese Wikipedia and 青空文庫 texts for POS-tagging and dependency-parsing (using `goeswith` for subwords), derived from [deberta-base-japanese-wikipedia](https://huggingface.co/KoichiYasuoka/deberta-base-japanese-wikipedia) and [UD_Japanese-GSDLUW](https://github.com/UniversalDependencies/UD_Japanese-GSDLUW).
+This is a DeBERTa(V2) model pretrained on Japanese Wikipedia and 青空文庫 texts for POS-tagging and dependency-parsing (using `goeswith` for subwords), derived from [deberta-base-japanese-wikipedia-luw-upos](https://huggingface.co/KoichiYasuoka/deberta-base-japanese-wikipedia-luw-upos) and [UD_Japanese-GSDLUW](https://github.com/UniversalDependencies/UD_Japanese-GSDLUW).
 
 ## How to Use
 

maker.py

@@ -1,5 +1,5 @@
 #! /usr/bin/python3
-src="KoichiYasuoka/deberta-base-japanese-wikipedia"
+src="KoichiYasuoka/deberta-base-japanese-wikipedia-luw-upos"
 tgt="KoichiYasuoka/deberta-base-japanese-wikipedia-ud-goeswith"
 url="https://github.com/UniversalDependencies/UD_Japanese-GSDLUW"
 import os
@@ -46,9 +46,9 @@ trainDS=UDgoeswithDataset("train.conllu",tkz)
 devDS=UDgoeswithDataset("dev.conllu",tkz)
 testDS=UDgoeswithDataset("test.conllu",tkz)
 lid=trainDS(devDS,testDS)
-cfg=AutoConfig.from_pretrained(src,num_labels=len(lid),label2id=lid,id2label={i:l for l,i in lid.items()})
+cfg=AutoConfig.from_pretrained(src,num_labels=len(lid),label2id=lid,id2label={i:l for l,i in lid.items()},ignore_mismatched_sizes=True,task_specific_params=None)
 arg=TrainingArguments(num_train_epochs=3,per_device_train_batch_size=32,output_dir="/tmp",overwrite_output_dir=True,save_total_limit=2,evaluation_strategy="epoch",learning_rate=5e-05,warmup_ratio=0.1)
-trn=Trainer(args=arg,data_collator=DataCollatorForTokenClassification(tkz),model=AutoModelForTokenClassification.from_pretrained(src,config=cfg),train_dataset=trainDS,eval_dataset=devDS)
+trn=Trainer(args=arg,data_collator=DataCollatorForTokenClassification(tkz),model=AutoModelForTokenClassification.from_pretrained(src,config=cfg,ignore_mismatched_sizes=True),train_dataset=trainDS,eval_dataset=devDS)
 trn.train()
 trn.save_model(tgt)
 tkz.save_pretrained(tgt)

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ec89b5bb67220bfb1cc206157a63327c0e7533501a7fb9e2fa7132a5dc7ef88e
+oid sha256:38da14abca09c2361a28518d060c6438e5bf3c3c99da1cf6584a69484d85b917
 size 498741331