model improved

Browse files

Files changed (6) hide show

README.md +1 -1
config.json +1 -1
maker.py +3 -3
pytorch_model.bin +2 -2
tokenizer_config.json +0 -1
vocab.txt +469 -0

README.md CHANGED Viewed

@@ -16,7 +16,7 @@ pipeline_tag: "token-classification"
 ## Model Description
-This is a RoBERTa model pre-trained on Chinese texts (both simplified and traditional) for POS-tagging and dependency-parsing (using `goeswith` for subwords), derived from [roberta_chinese_base](https://huggingface.co/clue/roberta_chinese_base).
 ## How to Use

 ## Model Description
+This is a RoBERTa model pre-trained on Chinese texts (both simplified and traditional) for POS-tagging and dependency-parsing (using `goeswith` for subwords), derived from [roberta-base-chinese-upos](https://huggingface.co/KoichiYasuoka/roberta-base-chinese-upos).
 ## How to Use

config.json CHANGED Viewed

@@ -644,5 +644,5 @@
   "transformers_version": "4.22.1",
   "type_vocab_size": 2,
   "use_cache": true,
-  "vocab_size": 21128
 }

   "transformers_version": "4.22.1",
   "type_vocab_size": 2,
   "use_cache": true,
+  "vocab_size": 21597
 }

maker.py CHANGED Viewed

@@ -1,5 +1,5 @@
 #! /usr/bin/python3
-src="clue/roberta_chinese_base"
 tgt="KoichiYasuoka/roberta-base-chinese-ud-goeswith"
 import os
 for d in ["UD_Chinese-GSD","UD_Chinese-GSDSimp"]:
@@ -45,9 +45,9 @@ 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()})
 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.train()
 trn.save_model(tgt)
 tkz.save_pretrained(tgt)

 #! /usr/bin/python3
+src="KoichiYasuoka/roberta-base-chinese-upos"
 tgt="KoichiYasuoka/roberta-base-chinese-ud-goeswith"
 import os
 for d in ["UD_Chinese-GSD","UD_Chinese-GSDSimp"]:
 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()},ignore_mismatched_sizes=True)
 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,ignore_mismatched_sizes=True),train_dataset=trainDS,eval_dataset=devDS)
 trn.train()
 trn.save_model(tgt)
 tkz.save_pretrained(tgt)

pytorch_model.bin CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dfb7b516490a2a111a771d186112e9a33f9aa9c39c54514a0db029382159fd45
-size 407711217

 version https://git-lfs.github.com/spec/v1
+oid sha256:d46ab28f34d65e8cd28c870bc9ca6e8d10d012e9a900df272620a292b67f5825
+size 409151985

tokenizer_config.json CHANGED Viewed

@@ -7,7 +7,6 @@
   "never_split": null,
   "pad_token": "[PAD]",
   "sep_token": "[SEP]",
-  "special_tokens_map_file": null,
   "strip_accents": null,
   "tokenize_chinese_chars": true,
   "tokenizer_class": "BertTokenizerFast",

   "never_split": null,
   "pad_token": "[PAD]",
   "sep_token": "[SEP]",
   "strip_accents": null,
   "tokenize_chinese_chars": true,
   "tokenizer_class": "BertTokenizerFast",

vocab.txt CHANGED Viewed

@@ -21126,3 +21126,472 @@ fishbase
 ##🔥
 ##😂
 ##😎

 ##🔥
 ##😂
 ##😎
+𫍟
+𧦧
+𰾵
+𨬟
+𫮃
+墠
+𪪼
+彃
+𢭏
+擣
+𥐟
+礒
+𬘝
+紾
+𫄨
+絺
+𮉪
+緅
+𰬸
+繐
+𫄸
+纁
+𮉡
+纑
+𫄥
+纚
+𦰏
+蓧
+𫉁
+薆
+訑
+𫍥
+誂
+𬤊
+諟
+𬤣
+譈
+𫐄
+軏
+𫐐
+輗
+𬨎
+輶
+𫓧
+鈇
+𨱂
+鈋
+𬱙
+頖
+𫖹
+顣
+𫗞
+飦
+𫗦
+餔
+𮩝
+餲
+𮩞
+饐
+𫗴
+饘
+𬶍
+鮀
+𫚈
+鱮
+𫛞
+鴃
+𫜁
+鷩
+㑹
+㧛
+㧞
+㨗
+㫖
+㱩
+殰
+䃅
+磾
+䌶
+䊷
+䌷
+紬
+䌹
+絅
+䘮
+䜣
+訢
+䟽
+䧟
+䯄
+騧
+䶮
+龑
+竝
+伥
+倀
+倂
+侮
+倹
+偸
+僂
+偾
+僨
+儺
+僞
+僧
+免
+兎
+勉
+勤
+勲
+卑
+卽
+厠
+廄
+嘆
+呉
+吿
+呑
+呙
+咼
+呪
+哙
+噲
+喩
+器
+嚢
+圹
+壙
+埀
+塡
+墨
+壌
+壱
+奨
+嫗
+妬
+娯
+嬋
+媭
+嬃
+賓
+尙
+尭
+屦
+屨
+巌
+巣
+帏
+幃
+帱
+幬
+幷
+廐
+廪
+廩
+廵
+忾
+愾
+怃
+憮
+愴
+懟
+恠
+惻
+悔
+懲
+愠
+慍
+愼
+憎
+懐
+戯
+扡
+拝
+挿
+揔
+摠
+摂
+摅
+攄
+擯
+撄
+攖
+擥
+擧
+攷
+敏
+斉
+旣
+暁
+暦
+朞
+殺
+枨
+棖
+柰
+橈
+梅
+梼
+檮
+棁
+梲
+椟
+櫝
+榅
+榲
+楡
+槚
+檟
+槨
+槪
+櫓
+欤
+歟
+歿
+漢
+汚
+浍
+澮
+海
+涜
+淸
+渇
+渓
+漑
+潅
+竈
+焭
+煢
+煕
+犠
+貍
+猟
+玆
+疴
+痾
+癰
+益
+著
+硁
+硜
+硗
+磽
+社
+祈
+祐
+祖
+祝
+神
+祥
+禎
+禍
+福
+秊
+穀
+穑
+穡
+穣
+穰
+突
+笾
+籩
+筚
+篳
+箪
+簞
+節
+簣
+簒
+籴
+糴
+類
+粛
+緖
+緜
+緼
+縕
+縦
+繁
+繊
+繍
+绐
+紿
+绖
+絰
+绤
+綌
+绹
+綯
+缁
+緇
+缊
+缌
+緦
+缗
+緡
+缧
+縲
+缫
+繅
+缵
+纘
+者
+聡
+臭
+茕
+荅
+荛
+蕘
+蒉
+蕢
+蔂
+虆
+薫
+虜
+虯
+蛍
+鼃
+螻
+蝿
+褐
+視
+觌
+覿
+覲
+觴
+訚
+誾
+諸
+謁
+謹
+讐
+讎
+讦
+訐
+讱
+訒
+謳
+诎
+詘
+诐
+詖
+诒
+詒
+诔
+誄
+诼
+諑
+諛
+諂
+谇
+誶
+讒
+謚
+謫
+谮
+譖
+賛
+贶
+貺
+賅
+赆
+贐
+赉
+賚
+踯
+躑
+轫
+軔
+軻
+辂
+輅
+辔
+轡
+逓
+逸
+郞
+都
+鄕
+酔
+醤
+釈
+鋭
+鋳
+錬
+鎭
+鑚
+鉞
+鑠
+锜
+錡
+锸
+鍤
+镃
+鎡
+镒
+鎰
+镦
+鐓
+鬭
+闋
+陥
+隷
+難
+雠
+靁
+靑
+韫
+韞
+頻
+顕
+顚
+顙
+飮
+餍
+饜
+饩
+餼
+飭
+馑
+饉
+騒
+駟
+駢
+驪
+骍
+騂
+騫
+鬪
+鲂
+魴
+鲐
+鮐
+鲧
+鯀
+鳏
+鰥
+鴈
+鶏
+鷄
+鸱
+鴟
+鸷
+鷙
+鵠
+鶻
+鹥
+鷖
+鹯
+鸇
+麪
+麹
+黙
+鼇
+鼈
+黿
+鼌
+鼂
+鼍
+鼉
+龁
+齕
+隆
+精
+羽
+飯
+館
+既