initial release

README.md

@@ -0,0 +1,74 @@
+---
+language:
+- "zh"
+tags:
+- "chinese"
+- "token-classification"
+- "pos"
+- "dependency-parsing"
+datasets:
+- "universal_dependencies"
+license: "apache-2.0"
+pipeline_tag: "token-classification"
+---
+
+# roberta-base-chinese-ud-goeswith
+
+## Model Description
+
+This is a RoBERTa model pre-trained on Chinese Wikipedia texts (both simplified and traditional) for POS-tagging and dependency-parsing (using `goeswith` for subwords), derived from [chinese-roberta-base-upos](https://huggingface.co/KoichiYasuoka/chinese-roberta-base-upos).
+
+## How to Use
+
+```py
+class UDgoeswith(object):
+  def __init__(self,bert):
+    from transformers import AutoTokenizer,AutoModelForTokenClassification
+    self.tokenizer=AutoTokenizer.from_pretrained(bert)
+    self.model=AutoModelForTokenClassification.from_pretrained(bert)
+  def __call__(self,text):
+    import numpy,torch,ufal.chu_liu_edmonds
+    w=self.tokenizer(text,return_offsets_mapping=True)
+    v=w["input_ids"]
+    x=[v[0:i]+[self.tokenizer.mask_token_id]+v[i+1:]+[j] for i,j in enumerate(v[1:-1],1)]
+    with torch.no_grad():
+      e=self.model(input_ids=torch.tensor(x)).logits.numpy()[:,1:-2,:]
+    r=[1 if i==0 else -1 if j.endswith("|root") else 0 for i,j in sorted(self.model.config.id2label.items())]
+    e+=numpy.where(numpy.add.outer(numpy.identity(e.shape[0]),r)==0,0,numpy.nan)
+    g=self.model.config.label2id["X|_|goeswith"]
+    r=numpy.tri(e.shape[0])
+    for i in range(e.shape[0]):
+      for j in range(i+2,e.shape[1]):
+        r[i,j]=r[i,j-1] if numpy.nanargmax(e[i,j-1])==g else 1
+    e[:,:,g]+=numpy.where(r==0,0,numpy.nan)
+    m=numpy.full((e.shape[0]+1,e.shape[1]+1),numpy.nan)
+    m[1:,1:]=numpy.nanmax(e,axis=2).transpose()
+    p=numpy.zeros(m.shape)
+    p[1:,1:]=numpy.nanargmax(e,axis=2).transpose()
+    for i in range(1,m.shape[0]):
+      m[i,0],m[i,i],p[i,0]=m[i,i],numpy.nan,p[i,i]
+    h=ufal.chu_liu_edmonds.chu_liu_edmonds(m)[0]
+    if [0 for i in h if i==0]!=[0]:
+      m[:,0]+=numpy.where(m[:,0]==numpy.nanmax(m[[i for i,j in enumerate(h) if j==0],0]),0,numpy.nan)
+      m[[i for i,j in enumerate(h) if j==0]]+=[0 if i==0 or j==0 else numpy.nan for i,j in enumerate(h)]
+      h=ufal.chu_liu_edmonds.chu_liu_edmonds(m)[0]
+    u="# text = "+text+"\n"
+    v=[(s,e) for s,e in w["offset_mapping"] if s<e]
+    for i,(s,e) in enumerate(v,1):
+      q=self.model.config.id2label[p[i,h[i]]].split("|")
+      u+="\t".join([str(i),text[s:e],"_",q[0],"_","|".join(q[1:-1]),str(h[i]),q[-1],"_","_" if i<len(v) and e<v[i][0] else "SpaceAfter=No"])+"\n"
+    return u+"\n"
+
+nlp=UDgoeswith("KoichiYasuoka/roberta-base-chinese-ud-goeswith")
+print(nlp("我叫萨拉，我住在伦敦。"))
+```
+
+with [ufal.chu-liu-edmonds](https://pypi.org/project/ufal.chu-liu-edmonds/).
+Or without ufal.chu-liu-edmonds:
+
+```
+from transformers import pipeline
+nlp=pipeline("universal-dependencies","KoichiYasuoka/roberta-base-chinese-ud-goeswith",trust_remote_code=True,aggregation_strategy="simple")
+print(nlp("我叫萨拉，我住在伦敦。"))
+```
+

config.json

@@ -0,0 +1,651 @@
+{
+  "architectures": [
+    "BertForTokenClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": null,
+  "custom_pipelines": {
+    "universal-dependencies": {
+      "impl": "ud.UniversalDependenciesPipeline"
+    }
+  },
+  "directionality": "bidi",
+  "eos_token_id": 2,
+  "finetuning_task": "ner",
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "id2label": {
+    "0": "-|_|dep",
+    "1": "ADJ|_|acl",
+    "2": "ADJ|_|acl:relcl",
+    "3": "ADJ|_|advcl",
+    "4": "ADJ|_|advmod",
+    "5": "ADJ|_|amod",
+    "6": "ADJ|_|appos",
+    "7": "ADJ|_|case",
+    "8": "ADJ|_|ccomp",
+    "9": "ADJ|_|compound",
+    "10": "ADJ|_|conj",
+    "11": "ADJ|_|cop",
+    "12": "ADJ|_|csubj",
+    "13": "ADJ|_|dislocated",
+    "14": "ADJ|_|nmod",
+    "15": "ADJ|_|nsubj",
+    "16": "ADJ|_|nsubj:pass",
+    "17": "ADJ|_|nummod",
+    "18": "ADJ|_|obj",
+    "19": "ADJ|_|obl",
+    "20": "ADJ|_|parataxis",
+    "21": "ADJ|_|root",
+    "22": "ADJ|_|xcomp",
+    "23": "ADP|Case=Gen|acl",
+    "24": "ADP|Voice=Cau|case",
+    "25": "ADP|_|acl",
+    "26": "ADP|_|acl:relcl",
+    "27": "ADP|_|advcl",
+    "28": "ADP|_|advmod",
+    "29": "ADP|_|amod",
+    "30": "ADP|_|appos",
+    "31": "ADP|_|case",
+    "32": "ADP|_|cc",
+    "33": "ADP|_|ccomp",
+    "34": "ADP|_|conj",
+    "35": "ADP|_|csubj",
+    "36": "ADP|_|det",
+    "37": "ADP|_|mark",
+    "38": "ADP|_|nmod:tmod",
+    "39": "ADP|_|obl",
+    "40": "ADP|_|parataxis",
+    "41": "ADP|_|root",
+    "42": "ADP|_|xcomp",
+    "43": "ADV|Polarity=Neg|advmod",
+    "44": "ADV|_|acl",
+    "45": "ADV|_|advcl",
+    "46": "ADV|_|advmod",
+    "47": "ADV|_|amod",
+    "48": "ADV|_|cc",
+    "49": "ADV|_|ccomp",
+    "50": "ADV|_|conj",
+    "51": "ADV|_|mark",
+    "52": "ADV|_|nmod:tmod",
+    "53": "ADV|_|parataxis",
+    "54": "ADV|_|root",
+    "55": "ADV|_|xcomp",
+    "56": "AUX|Aspect=Perf|aux",
+    "57": "AUX|Aspect=Prog|aux",
+    "58": "AUX|Voice=Pass|aux:pass",
+    "59": "AUX|_|acl:relcl",
+    "60": "AUX|_|aux",
+    "61": "AUX|_|ccomp",
+    "62": "AUX|_|conj",
+    "63": "AUX|_|cop",
+    "64": "AUX|_|root",
+    "65": "AUX|_|xcomp",
+    "66": "CCONJ|_|cc",
+    "67": "DET|_|acl",
+    "68": "DET|_|advmod",
+    "69": "DET|_|amod",
+    "70": "DET|_|case",
+    "71": "DET|_|conj",
+    "72": "DET|_|det",
+    "73": "DET|_|nmod",
+    "74": "DET|_|nmod:tmod",
+    "75": "DET|_|nsubj",
+    "76": "DET|_|obl",
+    "77": "NOUN|Number=Plur|nmod",
+    "78": "NOUN|Number=Plur|nsubj",
+    "79": "NOUN|_|acl",
+    "80": "NOUN|_|acl:relcl",
+    "81": "NOUN|_|advcl",
+    "82": "NOUN|_|advmod",
+    "83": "NOUN|_|amod",
+    "84": "NOUN|_|appos",
+    "85": "NOUN|_|case",
+    "86": "NOUN|_|ccomp",
+    "87": "NOUN|_|clf",
+    "88": "NOUN|_|compound",
+    "89": "NOUN|_|conj",
+    "90": "NOUN|_|csubj",
+    "91": "NOUN|_|dislocated",
+    "92": "NOUN|_|iobj",
+    "93": "NOUN|_|mark",
+    "94": "NOUN|_|mark:rel",
+    "95": "NOUN|_|nmod",
+    "96": "NOUN|_|nmod:tmod",
+    "97": "NOUN|_|nsubj",
+    "98": "NOUN|_|nsubj:pass",
+    "99": "NOUN|_|nummod",
+    "100": "NOUN|_|obj",
+    "101": "NOUN|_|obl",
+    "102": "NOUN|_|obl:patient",
+    "103": "NOUN|_|orphan",
+    "104": "NOUN|_|parataxis",
+    "105": "NOUN|_|root",
+    "106": "NOUN|_|xcomp",
+    "107": "NUM|NumType=Card|acl",
+    "108": "NUM|NumType=Card|advcl",
+    "109": "NUM|NumType=Card|advmod",
+    "110": "NUM|NumType=Card|amod",
+    "111": "NUM|NumType=Card|appos",
+    "112": "NUM|NumType=Card|ccomp",
+    "113": "NUM|NumType=Card|compound",
+    "114": "NUM|NumType=Card|conj",
+    "115": "NUM|NumType=Card|nmod",
+    "116": "NUM|NumType=Card|nmod:tmod",
+    "117": "NUM|NumType=Card|nsubj",
+    "118": "NUM|NumType=Card|nsubj:pass",
+    "119": "NUM|NumType=Card|nummod",
+    "120": "NUM|NumType=Card|obj",
+    "121": "NUM|NumType=Card|obl",
+    "122": "NUM|NumType=Card|parataxis",
+    "123": "NUM|NumType=Card|root",
+    "124": "NUM|NumType=Card|xcomp",
+    "125": "NUM|NumType=Ord|advmod",
+    "126": "NUM|NumType=Ord|compound",
+    "127": "NUM|NumType=Ord|conj",
+    "128": "NUM|NumType=Ord|nmod",
+    "129": "NUM|NumType=Ord|nummod",
+    "130": "NUM|NumType=Ord|obj",
+    "131": "NUM|NumType=Ord|root",
+    "132": "NUM|_|nummod",
+    "133": "PART|Aspect=Perf|discourse",
+    "134": "PART|Aspect=Perf|parataxis",
+    "135": "PART|Case=Gen|case",
+    "136": "PART|Number=Plur|appos",
+    "137": "PART|Number=Plur|conj",
+    "138": "PART|Number=Plur|nmod",
+    "139": "PART|Number=Plur|nsubj",
+    "140": "PART|Number=Plur|obl",
+    "141": "PART|PartType=Int|discourse:sp",
+    "142": "PART|_|acl",
+    "143": "PART|_|acl:relcl",
+    "144": "PART|_|advcl",
+    "145": "PART|_|advmod",
+    "146": "PART|_|amod",
+    "147": "PART|_|appos",
+    "148": "PART|_|case",
+    "149": "PART|_|ccomp",
+    "150": "PART|_|compound",
+    "151": "PART|_|compound:ext",
+    "152": "PART|_|conj",
+    "153": "PART|_|csubj",
+    "154": "PART|_|discourse",
+    "155": "PART|_|iobj",
+    "156": "PART|_|mark:adv",
+    "157": "PART|_|mark:rel",
+    "158": "PART|_|nmod",
+    "159": "PART|_|nmod:tmod",
+    "160": "PART|_|nsubj",
+    "161": "PART|_|nsubj:pass",
+    "162": "PART|_|obj",
+    "163": "PART|_|obl",
+    "164": "PART|_|obl:patient",
+    "165": "PART|_|parataxis",
+    "166": "PART|_|root",
+    "167": "PART|_|xcomp",
+    "168": "PRON|Number=Plur|Person=1|det",
+    "169": "PRON|Number=Plur|Person=1|nmod",
+    "170": "PRON|Number=Plur|Person=1|nsubj",
+    "171": "PRON|Number=Plur|Person=1|obj",
+    "172": "PRON|Number=Plur|Person=1|obl",
+    "173": "PRON|Number=Plur|Person=3|appos",
+    "174": "PRON|Number=Plur|Person=3|det",
+    "175": "PRON|Number=Plur|Person=3|nmod",
+    "176": "PRON|Number=Plur|Person=3|nsubj",
+    "177": "PRON|Number=Plur|Person=3|nsubj:pass",
+    "178": "PRON|Number=Plur|Person=3|obj",
+    "179": "PRON|Number=Plur|Person=3|obl",
+    "180": "PRON|Number=Plur|Person=3|obl:patient",
+    "181": "PRON|Person=1|compound",
+    "182": "PRON|Person=1|det",
+    "183": "PRON|Person=1|nmod",
+    "184": "PRON|Person=1|nsubj",
+    "185": "PRON|Person=1|obj",
+    "186": "PRON|Person=2|det",
+    "187": "PRON|Person=2|iobj",
+    "188": "PRON|Person=2|nmod",
+    "189": "PRON|Person=2|nsubj",
+    "190": "PRON|Person=2|obj",
+    "191": "PRON|Person=2|obl",
+    "192": "PRON|Person=3|conj",
+    "193": "PRON|Person=3|det",
+    "194": "PRON|Person=3|iobj",
+    "195": "PRON|Person=3|nmod",
+    "196": "PRON|Person=3|nsubj",
+    "197": "PRON|Person=3|nsubj:pass",
+    "198": "PRON|Person=3|obj",
+    "199": "PRON|Person=3|obl",
+    "200": "PRON|Person=3|obl:patient",
+    "201": "PRON|Person=3|parataxis",
+    "202": "PRON|Person=3|xcomp",
+    "203": "PRON|_|advmod",
+    "204": "PRON|_|appos",
+    "205": "PRON|_|compound",
+    "206": "PRON|_|conj",
+    "207": "PRON|_|csubj",
+    "208": "PRON|_|det",
+    "209": "PRON|_|nmod",
+    "210": "PRON|_|nsubj",
+    "211": "PRON|_|nsubj:pass",
+    "212": "PRON|_|obj",
+    "213": "PRON|_|obl",
+    "214": "PRON|_|obl:patient",
+    "215": "PRON|_|parataxis",
+    "216": "PROPN|_|acl:relcl",
+    "217": "PROPN|_|advcl",
+    "218": "PROPN|_|advmod",
+    "219": "PROPN|_|appos",
+    "220": "PROPN|_|case",
+    "221": "PROPN|_|ccomp",
+    "222": "PROPN|_|compound",
+    "223": "PROPN|_|conj",
+    "224": "PROPN|_|dislocated",
+    "225": "PROPN|_|flat:foreign",
+    "226": "PROPN|_|flat:name",
+    "227": "PROPN|_|iobj",
+    "228": "PROPN|_|nmod",
+    "229": "PROPN|_|nmod:tmod",
+    "230": "PROPN|_|nsubj",
+    "231": "PROPN|_|nsubj:pass",
+    "232": "PROPN|_|nummod",
+    "233": "PROPN|_|obj",
+    "234": "PROPN|_|obl",
+    "235": "PROPN|_|obl:patient",
+    "236": "PROPN|_|parataxis",
+    "237": "PROPN|_|root",
+    "238": "PROPN|_|vocative",
+    "239": "PROPN|_|xcomp",
+    "240": "PUNCT|_|appos",
+    "241": "PUNCT|_|punct",
+    "242": "PUNCT|_|reparandum",
+    "243": "SYM|_|appos",
+    "244": "SYM|_|nsubj",
+    "245": "SYM|_|obj",
+    "246": "SYM|_|punct",
+    "247": "VERB|Voice=Cau|acl",
+    "248": "VERB|Voice=Cau|acl:relcl",
+    "249": "VERB|Voice=Cau|advcl",
+    "250": "VERB|Voice=Cau|amod",
+    "251": "VERB|Voice=Cau|aux:pass",
+    "252": "VERB|Voice=Cau|ccomp",
+    "253": "VERB|Voice=Cau|compound",
+    "254": "VERB|Voice=Cau|conj",
+    "255": "VERB|Voice=Cau|csubj",
+    "256": "VERB|Voice=Cau|mark",
+    "257": "VERB|Voice=Cau|parataxis",
+    "258": "VERB|Voice=Cau|root",
+    "259": "VERB|Voice=Cau|xcomp",
+    "260": "VERB|Voice=Pass|aux:pass",
+    "261": "VERB|_|acl",
+    "262": "VERB|_|acl:relcl",
+    "263": "VERB|_|advcl",
+    "264": "VERB|_|advmod",
+    "265": "VERB|_|amod",
+    "266": "VERB|_|appos",
+    "267": "VERB|_|ccomp",
+    "268": "VERB|_|compound",
+    "269": "VERB|_|conj",
+    "270": "VERB|_|csubj",
+    "271": "VERB|_|csubj:pass",
+    "272": "VERB|_|det",
+    "273": "VERB|_|discourse",
+    "274": "VERB|_|dislocated",
+    "275": "VERB|_|mark",
+    "276": "VERB|_|nmod",
+    "277": "VERB|_|nmod:tmod",
+    "278": "VERB|_|nsubj",
+    "279": "VERB|_|obj",
+    "280": "VERB|_|obl",
+    "281": "VERB|_|parataxis",
+    "282": "VERB|_|reparandum",
+    "283": "VERB|_|root",
+    "284": "VERB|_|xcomp",
+    "285": "X|_|acl",
+    "286": "X|_|advcl",
+    "287": "X|_|advmod",
+    "288": "X|_|amod",
+    "289": "X|_|appos",
+    "290": "X|_|ccomp",
+    "291": "X|_|compound",
+    "292": "X|_|conj",
+    "293": "X|_|det",
+    "294": "X|_|flat:foreign",
+    "295": "X|_|goeswith",
+    "296": "X|_|nmod",
+    "297": "X|_|nsubj",
+    "298": "X|_|nummod",
+    "299": "X|_|obj",
+    "300": "X|_|obl",
+    "301": "X|_|parataxis",
+    "302": "X|_|root"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "label2id": {
+    "-|_|dep": 0,
+    "ADJ|_|acl": 1,
+    "ADJ|_|acl:relcl": 2,
+    "ADJ|_|advcl": 3,
+    "ADJ|_|advmod": 4,
+    "ADJ|_|amod": 5,
+    "ADJ|_|appos": 6,
+    "ADJ|_|case": 7,
+    "ADJ|_|ccomp": 8,
+    "ADJ|_|compound": 9,
+    "ADJ|_|conj": 10,
+    "ADJ|_|cop": 11,
+    "ADJ|_|csubj": 12,
+    "ADJ|_|dislocated": 13,
+    "ADJ|_|nmod": 14,
+    "ADJ|_|nsubj": 15,
+    "ADJ|_|nsubj:pass": 16,
+    "ADJ|_|nummod": 17,
+    "ADJ|_|obj": 18,
+    "ADJ|_|obl": 19,
+    "ADJ|_|parataxis": 20,
+    "ADJ|_|root": 21,
+    "ADJ|_|xcomp": 22,
+    "ADP|Case=Gen|acl": 23,
+    "ADP|Voice=Cau|case": 24,
+    "ADP|_|acl": 25,
+    "ADP|_|acl:relcl": 26,
+    "ADP|_|advcl": 27,
+    "ADP|_|advmod": 28,
+    "ADP|_|amod": 29,
+    "ADP|_|appos": 30,
+    "ADP|_|case": 31,
+    "ADP|_|cc": 32,
+    "ADP|_|ccomp": 33,
+    "ADP|_|conj": 34,
+    "ADP|_|csubj": 35,
+    "ADP|_|det": 36,
+    "ADP|_|mark": 37,
+    "ADP|_|nmod:tmod": 38,
+    "ADP|_|obl": 39,
+    "ADP|_|parataxis": 40,
+    "ADP|_|root": 41,
+    "ADP|_|xcomp": 42,
+    "ADV|Polarity=Neg|advmod": 43,
+    "ADV|_|acl": 44,
+    "ADV|_|advcl": 45,
+    "ADV|_|advmod": 46,
+    "ADV|_|amod": 47,
+    "ADV|_|cc": 48,
+    "ADV|_|ccomp": 49,
+    "ADV|_|conj": 50,
+    "ADV|_|mark": 51,
+    "ADV|_|nmod:tmod": 52,
+    "ADV|_|parataxis": 53,
+    "ADV|_|root": 54,
+    "ADV|_|xcomp": 55,
+    "AUX|Aspect=Perf|aux": 56,
+    "AUX|Aspect=Prog|aux": 57,
+    "AUX|Voice=Pass|aux:pass": 58,
+    "AUX|_|acl:relcl": 59,
+    "AUX|_|aux": 60,
+    "AUX|_|ccomp": 61,
+    "AUX|_|conj": 62,
+    "AUX|_|cop": 63,
+    "AUX|_|root": 64,
+    "AUX|_|xcomp": 65,
+    "CCONJ|_|cc": 66,
+    "DET|_|acl": 67,
+    "DET|_|advmod": 68,
+    "DET|_|amod": 69,
+    "DET|_|case": 70,
+    "DET|_|conj": 71,
+    "DET|_|det": 72,
+    "DET|_|nmod": 73,
+    "DET|_|nmod:tmod": 74,
+    "DET|_|nsubj": 75,
+    "DET|_|obl": 76,
+    "NOUN|Number=Plur|nmod": 77,
+    "NOUN|Number=Plur|nsubj": 78,
+    "NOUN|_|acl": 79,
+    "NOUN|_|acl:relcl": 80,
+    "NOUN|_|advcl": 81,
+    "NOUN|_|advmod": 82,
+    "NOUN|_|amod": 83,
+    "NOUN|_|appos": 84,
+    "NOUN|_|case": 85,
+    "NOUN|_|ccomp": 86,
+    "NOUN|_|clf": 87,
+    "NOUN|_|compound": 88,
+    "NOUN|_|conj": 89,
+    "NOUN|_|csubj": 90,
+    "NOUN|_|dislocated": 91,
+    "NOUN|_|iobj": 92,
+    "NOUN|_|mark": 93,
+    "NOUN|_|mark:rel": 94,
+    "NOUN|_|nmod": 95,
+    "NOUN|_|nmod:tmod": 96,
+    "NOUN|_|nsubj": 97,
+    "NOUN|_|nsubj:pass": 98,
+    "NOUN|_|nummod": 99,
+    "NOUN|_|obj": 100,
+    "NOUN|_|obl": 101,
+    "NOUN|_|obl:patient": 102,
+    "NOUN|_|orphan": 103,
+    "NOUN|_|parataxis": 104,
+    "NOUN|_|root": 105,
+    "NOUN|_|xcomp": 106,
+    "NUM|NumType=Card|acl": 107,
+    "NUM|NumType=Card|advcl": 108,
+    "NUM|NumType=Card|advmod": 109,
+    "NUM|NumType=Card|amod": 110,
+    "NUM|NumType=Card|appos": 111,
+    "NUM|NumType=Card|ccomp": 112,
+    "NUM|NumType=Card|compound": 113,
+    "NUM|NumType=Card|conj": 114,
+    "NUM|NumType=Card|nmod": 115,
+    "NUM|NumType=Card|nmod:tmod": 116,
+    "NUM|NumType=Card|nsubj": 117,
+    "NUM|NumType=Card|nsubj:pass": 118,
+    "NUM|NumType=Card|nummod": 119,
+    "NUM|NumType=Card|obj": 120,
+    "NUM|NumType=Card|obl": 121,
+    "NUM|NumType=Card|parataxis": 122,
+    "NUM|NumType=Card|root": 123,
+    "NUM|NumType=Card|xcomp": 124,
+    "NUM|NumType=Ord|advmod": 125,
+    "NUM|NumType=Ord|compound": 126,
+    "NUM|NumType=Ord|conj": 127,
+    "NUM|NumType=Ord|nmod": 128,
+    "NUM|NumType=Ord|nummod": 129,
+    "NUM|NumType=Ord|obj": 130,
+    "NUM|NumType=Ord|root": 131,
+    "NUM|_|nummod": 132,
+    "PART|Aspect=Perf|discourse": 133,
+    "PART|Aspect=Perf|parataxis": 134,
+    "PART|Case=Gen|case": 135,
+    "PART|Number=Plur|appos": 136,
+    "PART|Number=Plur|conj": 137,
+    "PART|Number=Plur|nmod": 138,
+    "PART|Number=Plur|nsubj": 139,
+    "PART|Number=Plur|obl": 140,
+    "PART|PartType=Int|discourse:sp": 141,
+    "PART|_|acl": 142,
+    "PART|_|acl:relcl": 143,
+    "PART|_|advcl": 144,
+    "PART|_|advmod": 145,
+    "PART|_|amod": 146,
+    "PART|_|appos": 147,
+    "PART|_|case": 148,
+    "PART|_|ccomp": 149,
+    "PART|_|compound": 150,
+    "PART|_|compound:ext": 151,
+    "PART|_|conj": 152,
+    "PART|_|csubj": 153,
+    "PART|_|discourse": 154,
+    "PART|_|iobj": 155,
+    "PART|_|mark:adv": 156,
+    "PART|_|mark:rel": 157,
+    "PART|_|nmod": 158,
+    "PART|_|nmod:tmod": 159,
+    "PART|_|nsubj": 160,
+    "PART|_|nsubj:pass": 161,
+    "PART|_|obj": 162,
+    "PART|_|obl": 163,
+    "PART|_|obl:patient": 164,
+    "PART|_|parataxis": 165,
+    "PART|_|root": 166,
+    "PART|_|xcomp": 167,
+    "PRON|Number=Plur|Person=1|det": 168,
+    "PRON|Number=Plur|Person=1|nmod": 169,
+    "PRON|Number=Plur|Person=1|nsubj": 170,
+    "PRON|Number=Plur|Person=1|obj": 171,
+    "PRON|Number=Plur|Person=1|obl": 172,
+    "PRON|Number=Plur|Person=3|appos": 173,
+    "PRON|Number=Plur|Person=3|det": 174,
+    "PRON|Number=Plur|Person=3|nmod": 175,
+    "PRON|Number=Plur|Person=3|nsubj": 176,
+    "PRON|Number=Plur|Person=3|nsubj:pass": 177,
+    "PRON|Number=Plur|Person=3|obj": 178,
+    "PRON|Number=Plur|Person=3|obl": 179,
+    "PRON|Number=Plur|Person=3|obl:patient": 180,
+    "PRON|Person=1|compound": 181,
+    "PRON|Person=1|det": 182,
+    "PRON|Person=1|nmod": 183,
+    "PRON|Person=1|nsubj": 184,
+    "PRON|Person=1|obj": 185,
+    "PRON|Person=2|det": 186,
+    "PRON|Person=2|iobj": 187,
+    "PRON|Person=2|nmod": 188,
+    "PRON|Person=2|nsubj": 189,
+    "PRON|Person=2|obj": 190,
+    "PRON|Person=2|obl": 191,
+    "PRON|Person=3|conj": 192,
+    "PRON|Person=3|det": 193,
+    "PRON|Person=3|iobj": 194,
+    "PRON|Person=3|nmod": 195,
+    "PRON|Person=3|nsubj": 196,
+    "PRON|Person=3|nsubj:pass": 197,
+    "PRON|Person=3|obj": 198,
+    "PRON|Person=3|obl": 199,
+    "PRON|Person=3|obl:patient": 200,
+    "PRON|Person=3|parataxis": 201,
+    "PRON|Person=3|xcomp": 202,
+    "PRON|_|advmod": 203,
+    "PRON|_|appos": 204,
+    "PRON|_|compound": 205,
+    "PRON|_|conj": 206,
+    "PRON|_|csubj": 207,
+    "PRON|_|det": 208,
+    "PRON|_|nmod": 209,
+    "PRON|_|nsubj": 210,
+    "PRON|_|nsubj:pass": 211,
+    "PRON|_|obj": 212,
+    "PRON|_|obl": 213,
+    "PRON|_|obl:patient": 214,
+    "PRON|_|parataxis": 215,
+    "PROPN|_|acl:relcl": 216,
+    "PROPN|_|advcl": 217,
+    "PROPN|_|advmod": 218,
+    "PROPN|_|appos": 219,
+    "PROPN|_|case": 220,
+    "PROPN|_|ccomp": 221,
+    "PROPN|_|compound": 222,
+    "PROPN|_|conj": 223,
+    "PROPN|_|dislocated": 224,
+    "PROPN|_|flat:foreign": 225,
+    "PROPN|_|flat:name": 226,
+    "PROPN|_|iobj": 227,
+    "PROPN|_|nmod": 228,
+    "PROPN|_|nmod:tmod": 229,
+    "PROPN|_|nsubj": 230,
+    "PROPN|_|nsubj:pass": 231,
+    "PROPN|_|nummod": 232,
+    "PROPN|_|obj": 233,
+    "PROPN|_|obl": 234,
+    "PROPN|_|obl:patient": 235,
+    "PROPN|_|parataxis": 236,
+    "PROPN|_|root": 237,
+    "PROPN|_|vocative": 238,
+    "PROPN|_|xcomp": 239,
+    "PUNCT|_|appos": 240,
+    "PUNCT|_|punct": 241,
+    "PUNCT|_|reparandum": 242,
+    "SYM|_|appos": 243,
+    "SYM|_|nsubj": 244,
+    "SYM|_|obj": 245,
+    "SYM|_|punct": 246,
+    "VERB|Voice=Cau|acl": 247,
+    "VERB|Voice=Cau|acl:relcl": 248,
+    "VERB|Voice=Cau|advcl": 249,
+    "VERB|Voice=Cau|amod": 250,
+    "VERB|Voice=Cau|aux:pass": 251,
+    "VERB|Voice=Cau|ccomp": 252,
+    "VERB|Voice=Cau|compound": 253,
+    "VERB|Voice=Cau|conj": 254,
+    "VERB|Voice=Cau|csubj": 255,
+    "VERB|Voice=Cau|mark": 256,
+    "VERB|Voice=Cau|parataxis": 257,
+    "VERB|Voice=Cau|root": 258,
+    "VERB|Voice=Cau|xcomp": 259,
+    "VERB|Voice=Pass|aux:pass": 260,
+    "VERB|_|acl": 261,
+    "VERB|_|acl:relcl": 262,
+    "VERB|_|advcl": 263,
+    "VERB|_|advmod": 264,
+    "VERB|_|amod": 265,
+    "VERB|_|appos": 266,
+    "VERB|_|ccomp": 267,
+    "VERB|_|compound": 268,
+    "VERB|_|conj": 269,
+    "VERB|_|csubj": 270,
+    "VERB|_|csubj:pass": 271,
+    "VERB|_|det": 272,
+    "VERB|_|discourse": 273,
+    "VERB|_|dislocated": 274,
+    "VERB|_|mark": 275,
+    "VERB|_|nmod": 276,
+    "VERB|_|nmod:tmod": 277,
+    "VERB|_|nsubj": 278,
+    "VERB|_|obj": 279,
+    "VERB|_|obl": 280,
+    "VERB|_|parataxis": 281,
+    "VERB|_|reparandum": 282,
+    "VERB|_|root": 283,
+    "VERB|_|xcomp": 284,
+    "X|_|acl": 285,
+    "X|_|advcl": 286,
+    "X|_|advmod": 287,
+    "X|_|amod": 288,
+    "X|_|appos": 289,
+    "X|_|ccomp": 290,
+    "X|_|compound": 291,
+    "X|_|conj": 292,
+    "X|_|det": 293,
+    "X|_|flat:foreign": 294,
+    "X|_|goeswith": 295,
+    "X|_|nmod": 296,
+    "X|_|nsubj": 297,
+    "X|_|nummod": 298,
+    "X|_|obj": 299,
+    "X|_|obl": 300,
+    "X|_|parataxis": 301,
+    "X|_|root": 302
+  },
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "output_past": true,
+  "pad_token_id": 1,
+  "pooler_fc_size": 768,
+  "pooler_num_attention_heads": 12,
+  "pooler_num_fc_layers": 3,
+  "pooler_size_per_head": 128,
+  "pooler_type": "first_token_transform",
+  "position_embedding_type": "absolute",
+  "tokenizer_class": "BertTokenizerFast",
+  "torch_dtype": "float32",
+  "transformers_version": "4.22.1",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 21128
+}

maker.py

@@ -0,0 +1,53 @@
+#! /usr/bin/python3
+src="KoichiYasuoka/chinese-roberta-base-upos"
+tgt="KoichiYasuoka/roberta-base-chinese-ud-goeswith"
+import os
+for d in ["UD_Chinese-GSD","UD_Chinese-GSDSimp"]:
+  os.system("test -d "+d+" || git clone --depth=1 https://github.com/UniversalDependencies/"+d)
+os.system("for F in train dev test ; do cat UD_Chinese-*/*-$F.conllu > $F.conllu ; done")
+class UDgoeswithDataset(object):
+  def __init__(self,conllu,tokenizer):
+    self.ids,self.tags,label=[],[],set()
+    with open(conllu,"r",encoding="utf-8") as r:
+      cls,sep,msk=tokenizer.cls_token_id,tokenizer.sep_token_id,tokenizer.mask_token_id
+      dep,c="-|_|dep",[]
+      for s in r:
+        t=s.split("\t")
+        if len(t)==10 and t[0].isdecimal():
+          c.append(t)
+        elif c!=[]:
+          v=tokenizer([t[1] for t in c],add_special_tokens=False)["input_ids"]
+          for i in range(len(v)-1,-1,-1):
+            for j in range(1,len(v[i])):
+              c.insert(i+1,[c[i][0],"_","_","X","_","_",c[i][0],"goeswith","_","_"])
+          y=["0"]+[t[0] for t in c]
+          h=[i if t[6]=="0" else y.index(t[6]) for i,t in enumerate(c,1)]
+          p,v=[t[3]+"|"+t[5]+"|"+t[7] for t in c],sum(v,[])
+          self.ids.append([cls]+v+[sep])
+          self.tags.append([dep]+p+[dep])
+          label=set(sum([self.tags[-1],list(label)],[]))
+          for i,k in enumerate(v):
+            self.ids.append([cls]+v[0:i]+[msk]+v[i+1:]+[sep,k])
+            self.tags.append([dep]+[t if h[j]==i+1 else dep for j,t in enumerate(p)]+[dep,dep])
+          c=[]
+    self.label2id={l:i for i,l in enumerate(sorted(label))}
+  def __call__(*args):
+    label=set(sum([list(t.label2id) for t in args],[]))
+    lid={l:i for i,l in enumerate(sorted(label))}
+    for t in args:
+      t.label2id=lid
+    return lid
+  __len__=lambda self:len(self.ids)
+  __getitem__=lambda self,i:{"input_ids":self.ids[i],"labels":[self.label2id[t] for t in self.tags[i]]}
+from transformers import AutoTokenizer,AutoConfig,AutoModelForTokenClassification,DataCollatorForTokenClassification,TrainingArguments,Trainer
+tkz=AutoTokenizer.from_pretrained(src)
+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()},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

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:803211fef51e6bd23e68f211c1396d8975f16e2aea0ac0469a612fd83ccb38ba
+size 407710641

special_tokens_map.json

@@ -0,0 +1,7 @@
+{
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": "[UNK]"
+}

tokenizer_config.json

@@ -0,0 +1,14 @@
+{
+  "cls_token": "[CLS]",
+  "do_basic_tokenize": true,
+  "do_lower_case": true,
+  "mask_token": "[MASK]",
+  "model_max_length": 512,
+  "never_split": null,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "BertTokenizerFast",
+  "unk_token": "[UNK]"
+}

ud.py

@@ -0,0 +1,61 @@
+from transformers import TokenClassificationPipeline
+
+class UniversalDependenciesPipeline(TokenClassificationPipeline):
+  def _forward(self,model_input):
+    import torch
+    v=model_input["input_ids"][0].tolist()
+    with torch.no_grad():
+      e=self.model(input_ids=torch.tensor([v[0:i]+[self.tokenizer.mask_token_id]+v[i+1:]+[j] for i,j in enumerate(v[1:-1],1)]))
+    return {"logits":e.logits[:,1:-2,:],**model_input}
+  def postprocess(self,model_output,**kwargs):
+    import numpy
+    e=model_output["logits"].numpy()
+    r=[1 if i==0 else -1 if j.endswith("|root") else 0 for i,j in sorted(self.model.config.id2label.items())]
+    e+=numpy.where(numpy.add.outer(numpy.identity(e.shape[0]),r)==0,0,numpy.nan)
+    g=self.model.config.label2id["X|_|goeswith"]
+    r=numpy.tri(e.shape[0])
+    for i in range(e.shape[0]):
+      for j in range(i+2,e.shape[1]):
+        r[i,j]=r[i,j-1] if numpy.nanargmax(e[i,j-1])==g else 1
+    e[:,:,g]+=numpy.where(r==0,0,numpy.nan)
+    m,p=numpy.nanmax(e,axis=2),numpy.nanargmax(e,axis=2)
+    h=self.chu_liu_edmonds(m)
+    z=[i for i,j in enumerate(h) if i==j]
+    if len(z)>1:
+      k,h=z[numpy.nanargmax(m[z,z])],numpy.nanmin(m)-numpy.nanmax(m)
+      m[:,z]+=[[0 if j in z and (i!=j or i==k) else h for i in z] for j in range(m.shape[0])]
+      h=self.chu_liu_edmonds(m)
+    v=[(s,e) for s,e in model_output["offset_mapping"][0].tolist() if s<e]
+    q=[self.model.config.id2label[p[j,i]].split("|") for i,j in enumerate(h)]
+    g="aggregation_strategy" in kwargs and kwargs["aggregation_strategy"]!="none"
+    if g:
+      for i,j in reversed(list(enumerate(q[1:],1))):
+        if j[-1]=="goeswith" and set([t[-1] for t in q[h[i]+1:i+1]])=={"goeswith"}:
+          h=[b if i>b else b-1 for a,b in enumerate(h) if i!=a]
+          v[i-1]=(v[i-1][0],v.pop(i)[1])
+          q.pop(i)
+    t=model_output["sentence"].replace("\n"," ")
+    u="# text = "+t+"\n"
+    for i,(s,e) in enumerate(v):
+      u+="\t".join([str(i+1),t[s:e],t[s:e] if g else "_",q[i][0],"_","|".join(q[i][1:-1]),str(0 if h[i]==i else h[i]+1),q[i][-1],"_","_" if i+1<len(v) and e<v[i+1][0] else "SpaceAfter=No"])+"\n"
+    return u+"\n"
+  def chu_liu_edmonds(self,matrix):
+    import numpy
+    h=numpy.nanargmax(matrix,axis=0)
+    x=[-1 if i==j else j for i,j in enumerate(h)]
+    for b in [lambda x,i,j:-1 if i not in x else x[i],lambda x,i,j:-1 if j<0 else x[j]]:
+      y=[]
+      while x!=y:
+        y=list(x)
+        for i,j in enumerate(x):
+          x[i]=b(x,i,j)
+      if max(x)<0:
+        return h
+    y,x=[i for i,j in enumerate(x) if j==max(x)],[i for i,j in enumerate(x) if j<max(x)]
+    z=matrix-numpy.nanmax(matrix,axis=0)
+    m=numpy.block([[z[x,:][:,x],numpy.nanmax(z[x,:][:,y],axis=1).reshape(len(x),1)],[numpy.nanmax(z[y,:][:,x],axis=0),numpy.nanmax(z[y,y])]])
+    k=[j if i==len(x) else x[j] if j<len(x) else y[numpy.nanargmax(z[y,x[i]])] for i,j in enumerate(self.chu_liu_edmonds(m))]
+    h=[j if i in y else k[x.index(i)] for i,j in enumerate(h)]
+    i=y[numpy.nanargmax(z[x[k[-1]],y] if k[-1]<len(x) else z[y,y])]
+    h[i]=x[k[-1]] if k[-1]<len(x) else i
+    return h