initial release

README.md

@@ -0,0 +1,76 @@
+---
+language:
+- "th"
+tags:
+- "thai"
+- "token-classification"
+- "pos"
+- "dependency-parsing"
+datasets:
+- "universal_dependencies"
+license: "apache-2.0"
+pipeline_tag: "token-classification"
+widget:
+- text: "หลายหัวดีกว่าหัวเดียว"
+---
+
+# roberta-base-thai-spm-ud-goeswith
+
+## Model Description
+
+This is a RoBERTa model pre-trained on Thai Wikipedia texts for POS-tagging and dependency-parsing (using `goeswith` for subwords), derived from [roberta-base-thai-spm](https://huggingface.co/KoichiYasuoka/roberta-base-thai-spm).
+
+## 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-thai-spm-ud-goeswith")
+print(nlp("หลายหัวดีกว่าหัวเดียว"))
+```
+
+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-thai-spm-ud-goeswith",trust_remote_code=True,aggregation_strategy="simple")
+print(nlp("หลายหัวดีกว่าหัวเดียว"))
+```
+

config.json

@@ -0,0 +1,436 @@
+{
+  "architectures": [
+    "RobertaForTokenClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": null,
+  "custom_pipelines": {
+    "universal-dependencies": {
+      "impl": "ud.UniversalDependenciesPipeline"
+    }
+  },
+  "eos_token_id": 2,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "id2label": {
+    "0": "-|_|dep",
+    "1": "ADP|_|acl",
+    "2": "ADP|_|advcl",
+    "3": "ADP|_|advmod",
+    "4": "ADP|_|appos",
+    "5": "ADP|_|case",
+    "6": "ADP|_|cc",
+    "7": "ADP|_|cc:preconj",
+    "8": "ADP|_|csubj",
+    "9": "ADP|_|fixed",
+    "10": "ADP|_|mark",
+    "11": "ADP|_|obl",
+    "12": "ADP|_|root",
+    "13": "ADV|_|advcl",
+    "14": "ADV|_|advmod",
+    "15": "ADV|_|aux",
+    "16": "ADV|_|cc",
+    "17": "ADV|_|ccomp",
+    "18": "ADV|_|conj",
+    "19": "ADV|_|fixed",
+    "20": "ADV|_|mark",
+    "21": "ADV|_|obj",
+    "22": "ADV|_|root",
+    "23": "ADV|_|xcomp",
+    "24": "AUX|_|advmod",
+    "25": "AUX|_|aux",
+    "26": "AUX|_|aux:pass",
+    "27": "AUX|_|ccomp",
+    "28": "AUX|_|conj",
+    "29": "AUX|_|cop",
+    "30": "AUX|_|mark",
+    "31": "CCONJ|_|advmod",
+    "32": "CCONJ|_|case",
+    "33": "CCONJ|_|cc",
+    "34": "CCONJ|_|compound",
+    "35": "CCONJ|_|conj",
+    "36": "CCONJ|_|fixed",
+    "37": "CCONJ|_|mark",
+    "38": "CCONJ|_|nsubj",
+    "39": "CCONJ|_|obl",
+    "40": "CCONJ|_|root",
+    "41": "DET|_|advmod",
+    "42": "DET|_|case",
+    "43": "DET|_|cc:preconj",
+    "44": "DET|_|conj",
+    "45": "DET|_|det",
+    "46": "DET|_|det:predet",
+    "47": "DET|_|fixed",
+    "48": "DET|_|mark",
+    "49": "DET|_|nsubj",
+    "50": "DET|_|nsubj:pass",
+    "51": "DET|_|obj",
+    "52": "DET|_|obl",
+    "53": "DET|_|obl:tmod",
+    "54": "DET|_|root",
+    "55": "INTJ|_|acl",
+    "56": "INTJ|_|nsubj",
+    "57": "INTJ|_|root",
+    "58": "NOUN|_|acl",
+    "59": "NOUN|_|acl:relcl",
+    "60": "NOUN|_|advcl",
+    "61": "NOUN|_|advmod",
+    "62": "NOUN|_|appos",
+    "63": "NOUN|_|aux",
+    "64": "NOUN|_|case",
+    "65": "NOUN|_|cc",
+    "66": "NOUN|_|ccomp",
+    "67": "NOUN|_|clf",
+    "68": "NOUN|_|compound",
+    "69": "NOUN|_|conj",
+    "70": "NOUN|_|dislocated",
+    "71": "NOUN|_|fixed",
+    "72": "NOUN|_|flat:name",
+    "73": "NOUN|_|iobj",
+    "74": "NOUN|_|mark",
+    "75": "NOUN|_|nmod",
+    "76": "NOUN|_|nmod:poss",
+    "77": "NOUN|_|nsubj",
+    "78": "NOUN|_|nsubj:pass",
+    "79": "NOUN|_|obj",
+    "80": "NOUN|_|obl",
+    "81": "NOUN|_|obl:poss",
+    "82": "NOUN|_|obl:tmod",
+    "83": "NOUN|_|parataxis",
+    "84": "NOUN|_|root",
+    "85": "NOUN|_|vocative",
+    "86": "NOUN|_|xcomp",
+    "87": "NUM|_|acl",
+    "88": "NUM|_|acl:relcl",
+    "89": "NUM|_|advmod",
+    "90": "NUM|_|appos",
+    "91": "NUM|_|ccomp",
+    "92": "NUM|_|clf",
+    "93": "NUM|_|conj",
+    "94": "NUM|_|flat:name",
+    "95": "NUM|_|nmod",
+    "96": "NUM|_|nsubj",
+    "97": "NUM|_|nummod",
+    "98": "NUM|_|obj",
+    "99": "NUM|_|obl",
+    "100": "NUM|_|obl:poss",
+    "101": "NUM|_|obl:tmod",
+    "102": "NUM|_|root",
+    "103": "NUM|_|xcomp",
+    "104": "PART|_|acl",
+    "105": "PART|_|advmod",
+    "106": "PART|_|aux",
+    "107": "PART|_|cc",
+    "108": "PART|_|cc:preconj",
+    "109": "PART|_|ccomp",
+    "110": "PART|_|clf",
+    "111": "PART|_|compound",
+    "112": "PART|_|compound:prt",
+    "113": "PART|_|conj",
+    "114": "PART|_|discourse",
+    "115": "PART|_|fixed",
+    "116": "PART|_|mark",
+    "117": "PART|_|nmod",
+    "118": "PART|_|nmod:poss",
+    "119": "PART|_|nsubj",
+    "120": "PART|_|obj",
+    "121": "PART|_|obl",
+    "122": "PART|_|root",
+    "123": "PART|_|xcomp",
+    "124": "PRON|_|acl",
+    "125": "PRON|_|acl:relcl",
+    "126": "PRON|_|advcl",
+    "127": "PRON|_|advmod",
+    "128": "PRON|_|appos",
+    "129": "PRON|_|ccomp",
+    "130": "PRON|_|compound",
+    "131": "PRON|_|conj",
+    "132": "PRON|_|fixed",
+    "133": "PRON|_|nmod",
+    "134": "PRON|_|nmod:poss",
+    "135": "PRON|_|nsubj",
+    "136": "PRON|_|nsubj:pass",
+    "137": "PRON|_|obj",
+    "138": "PRON|_|obl",
+    "139": "PRON|_|obl:poss",
+    "140": "PRON|_|reparandum",
+    "141": "PRON|_|root",
+    "142": "PRON|_|xcomp",
+    "143": "PROPN|_|acl",
+    "144": "PROPN|_|acl:relcl",
+    "145": "PROPN|_|advmod",
+    "146": "PROPN|_|appos",
+    "147": "PROPN|_|aux",
+    "148": "PROPN|_|cc",
+    "149": "PROPN|_|ccomp",
+    "150": "PROPN|_|clf",
+    "151": "PROPN|_|compound",
+    "152": "PROPN|_|conj",
+    "153": "PROPN|_|flat:name",
+    "154": "PROPN|_|goeswith",
+    "155": "PROPN|_|nmod",
+    "156": "PROPN|_|nmod:poss",
+    "157": "PROPN|_|nsubj",
+    "158": "PROPN|_|nsubj:pass",
+    "159": "PROPN|_|obj",
+    "160": "PROPN|_|obl",
+    "161": "PROPN|_|obl:poss",
+    "162": "PROPN|_|obl:tmod",
+    "163": "PROPN|_|root",
+    "164": "PROPN|_|xcomp",
+    "165": "PUNCT|_|advmod",
+    "166": "PUNCT|_|clf",
+    "167": "PUNCT|_|punct",
+    "168": "PUNCT|_|root",
+    "169": "SCONJ|_|mark",
+    "170": "SYM|_|advmod",
+    "171": "SYM|_|clf",
+    "172": "SYM|_|nsubj",
+    "173": "SYM|_|obj",
+    "174": "SYM|_|obl",
+    "175": "VERB|_|acl",
+    "176": "VERB|_|acl:relcl",
+    "177": "VERB|_|advcl",
+    "178": "VERB|_|advmod",
+    "179": "VERB|_|appos",
+    "180": "VERB|_|aux",
+    "181": "VERB|_|aux:pass",
+    "182": "VERB|_|case",
+    "183": "VERB|_|cc",
+    "184": "VERB|_|ccomp",
+    "185": "VERB|_|compound",
+    "186": "VERB|_|conj",
+    "187": "VERB|_|csubj",
+    "188": "VERB|_|fixed",
+    "189": "VERB|_|mark",
+    "190": "VERB|_|nmod",
+    "191": "VERB|_|nmod:poss",
+    "192": "VERB|_|nsubj",
+    "193": "VERB|_|obj",
+    "194": "VERB|_|obl",
+    "195": "VERB|_|obl:poss",
+    "196": "VERB|_|parataxis",
+    "197": "VERB|_|root",
+    "198": "VERB|_|xcomp",
+    "199": "X|_|goeswith"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "label2id": {
+    "-|_|dep": 0,
+    "ADP|_|acl": 1,
+    "ADP|_|advcl": 2,
+    "ADP|_|advmod": 3,
+    "ADP|_|appos": 4,
+    "ADP|_|case": 5,
+    "ADP|_|cc": 6,
+    "ADP|_|cc:preconj": 7,
+    "ADP|_|csubj": 8,
+    "ADP|_|fixed": 9,
+    "ADP|_|mark": 10,
+    "ADP|_|obl": 11,
+    "ADP|_|root": 12,
+    "ADV|_|advcl": 13,
+    "ADV|_|advmod": 14,
+    "ADV|_|aux": 15,
+    "ADV|_|cc": 16,
+    "ADV|_|ccomp": 17,
+    "ADV|_|conj": 18,
+    "ADV|_|fixed": 19,
+    "ADV|_|mark": 20,
+    "ADV|_|obj": 21,
+    "ADV|_|root": 22,
+    "ADV|_|xcomp": 23,
+    "AUX|_|advmod": 24,
+    "AUX|_|aux": 25,
+    "AUX|_|aux:pass": 26,
+    "AUX|_|ccomp": 27,
+    "AUX|_|conj": 28,
+    "AUX|_|cop": 29,
+    "AUX|_|mark": 30,
+    "CCONJ|_|advmod": 31,
+    "CCONJ|_|case": 32,
+    "CCONJ|_|cc": 33,
+    "CCONJ|_|compound": 34,
+    "CCONJ|_|conj": 35,
+    "CCONJ|_|fixed": 36,
+    "CCONJ|_|mark": 37,
+    "CCONJ|_|nsubj": 38,
+    "CCONJ|_|obl": 39,
+    "CCONJ|_|root": 40,
+    "DET|_|advmod": 41,
+    "DET|_|case": 42,
+    "DET|_|cc:preconj": 43,
+    "DET|_|conj": 44,
+    "DET|_|det": 45,
+    "DET|_|det:predet": 46,
+    "DET|_|fixed": 47,
+    "DET|_|mark": 48,
+    "DET|_|nsubj": 49,
+    "DET|_|nsubj:pass": 50,
+    "DET|_|obj": 51,
+    "DET|_|obl": 52,
+    "DET|_|obl:tmod": 53,
+    "DET|_|root": 54,
+    "INTJ|_|acl": 55,
+    "INTJ|_|nsubj": 56,
+    "INTJ|_|root": 57,
+    "NOUN|_|acl": 58,
+    "NOUN|_|acl:relcl": 59,
+    "NOUN|_|advcl": 60,
+    "NOUN|_|advmod": 61,
+    "NOUN|_|appos": 62,
+    "NOUN|_|aux": 63,
+    "NOUN|_|case": 64,
+    "NOUN|_|cc": 65,
+    "NOUN|_|ccomp": 66,
+    "NOUN|_|clf": 67,
+    "NOUN|_|compound": 68,
+    "NOUN|_|conj": 69,
+    "NOUN|_|dislocated": 70,
+    "NOUN|_|fixed": 71,
+    "NOUN|_|flat:name": 72,
+    "NOUN|_|iobj": 73,
+    "NOUN|_|mark": 74,
+    "NOUN|_|nmod": 75,
+    "NOUN|_|nmod:poss": 76,
+    "NOUN|_|nsubj": 77,
+    "NOUN|_|nsubj:pass": 78,
+    "NOUN|_|obj": 79,
+    "NOUN|_|obl": 80,
+    "NOUN|_|obl:poss": 81,
+    "NOUN|_|obl:tmod": 82,
+    "NOUN|_|parataxis": 83,
+    "NOUN|_|root": 84,
+    "NOUN|_|vocative": 85,
+    "NOUN|_|xcomp": 86,
+    "NUM|_|acl": 87,
+    "NUM|_|acl:relcl": 88,
+    "NUM|_|advmod": 89,
+    "NUM|_|appos": 90,
+    "NUM|_|ccomp": 91,
+    "NUM|_|clf": 92,
+    "NUM|_|conj": 93,
+    "NUM|_|flat:name": 94,
+    "NUM|_|nmod": 95,
+    "NUM|_|nsubj": 96,
+    "NUM|_|nummod": 97,
+    "NUM|_|obj": 98,
+    "NUM|_|obl": 99,
+    "NUM|_|obl:poss": 100,
+    "NUM|_|obl:tmod": 101,
+    "NUM|_|root": 102,
+    "NUM|_|xcomp": 103,
+    "PART|_|acl": 104,
+    "PART|_|advmod": 105,
+    "PART|_|aux": 106,
+    "PART|_|cc": 107,
+    "PART|_|cc:preconj": 108,
+    "PART|_|ccomp": 109,
+    "PART|_|clf": 110,
+    "PART|_|compound": 111,
+    "PART|_|compound:prt": 112,
+    "PART|_|conj": 113,
+    "PART|_|discourse": 114,
+    "PART|_|fixed": 115,
+    "PART|_|mark": 116,
+    "PART|_|nmod": 117,
+    "PART|_|nmod:poss": 118,
+    "PART|_|nsubj": 119,
+    "PART|_|obj": 120,
+    "PART|_|obl": 121,
+    "PART|_|root": 122,
+    "PART|_|xcomp": 123,
+    "PRON|_|acl": 124,
+    "PRON|_|acl:relcl": 125,
+    "PRON|_|advcl": 126,
+    "PRON|_|advmod": 127,
+    "PRON|_|appos": 128,
+    "PRON|_|ccomp": 129,
+    "PRON|_|compound": 130,
+    "PRON|_|conj": 131,
+    "PRON|_|fixed": 132,
+    "PRON|_|nmod": 133,
+    "PRON|_|nmod:poss": 134,
+    "PRON|_|nsubj": 135,
+    "PRON|_|nsubj:pass": 136,
+    "PRON|_|obj": 137,
+    "PRON|_|obl": 138,
+    "PRON|_|obl:poss": 139,
+    "PRON|_|reparandum": 140,
+    "PRON|_|root": 141,
+    "PRON|_|xcomp": 142,
+    "PROPN|_|acl": 143,
+    "PROPN|_|acl:relcl": 144,
+    "PROPN|_|advmod": 145,
+    "PROPN|_|appos": 146,
+    "PROPN|_|aux": 147,
+    "PROPN|_|cc": 148,
+    "PROPN|_|ccomp": 149,
+    "PROPN|_|clf": 150,
+    "PROPN|_|compound": 151,
+    "PROPN|_|conj": 152,
+    "PROPN|_|flat:name": 153,
+    "PROPN|_|goeswith": 154,
+    "PROPN|_|nmod": 155,
+    "PROPN|_|nmod:poss": 156,
+    "PROPN|_|nsubj": 157,
+    "PROPN|_|nsubj:pass": 158,
+    "PROPN|_|obj": 159,
+    "PROPN|_|obl": 160,
+    "PROPN|_|obl:poss": 161,
+    "PROPN|_|obl:tmod": 162,
+    "PROPN|_|root": 163,
+    "PROPN|_|xcomp": 164,
+    "PUNCT|_|advmod": 165,
+    "PUNCT|_|clf": 166,
+    "PUNCT|_|punct": 167,
+    "PUNCT|_|root": 168,
+    "SCONJ|_|mark": 169,
+    "SYM|_|advmod": 170,
+    "SYM|_|clf": 171,
+    "SYM|_|nsubj": 172,
+    "SYM|_|obj": 173,
+    "SYM|_|obl": 174,
+    "VERB|_|acl": 175,
+    "VERB|_|acl:relcl": 176,
+    "VERB|_|advcl": 177,
+    "VERB|_|advmod": 178,
+    "VERB|_|appos": 179,
+    "VERB|_|aux": 180,
+    "VERB|_|aux:pass": 181,
+    "VERB|_|case": 182,
+    "VERB|_|cc": 183,
+    "VERB|_|ccomp": 184,
+    "VERB|_|compound": 185,
+    "VERB|_|conj": 186,
+    "VERB|_|csubj": 187,
+    "VERB|_|fixed": 188,
+    "VERB|_|mark": 189,
+    "VERB|_|nmod": 190,
+    "VERB|_|nmod:poss": 191,
+    "VERB|_|nsubj": 192,
+    "VERB|_|obj": 193,
+    "VERB|_|obl": 194,
+    "VERB|_|obl:poss": 195,
+    "VERB|_|parataxis": 196,
+    "VERB|_|root": 197,
+    "VERB|_|xcomp": 198,
+    "X|_|goeswith": 199
+  },
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "roberta",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 1,
+  "position_embedding_type": "absolute",
+  "tokenizer_class": "RemBertTokenizerFast",
+  "torch_dtype": "float32",
+  "transformers_version": "4.22.1",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 3005
+}

maker.py

@@ -0,0 +1,47 @@
+#! /usr/bin/python3
+src="KoichiYasuoka/roberta-base-thai-spm"
+tgt="KoichiYasuoka/roberta-base-thai-spm-ud-goeswith"
+url="https://github.com/KoichiYasuoka/spaCy-Thai"
+import os
+d=os.path.join(os.path.basename(url),"UD_Thai-Corpora")
+os.system("test -d {} || git clone --depth=1 {}".format(d,url))
+s='{if(NF>0)u=u$0"\\n";else{if(u~/\\t0\\troot\\t/)print u>"train.conllu";u=""}}'
+os.system("nawk -F'\\t' '{}' {}/*-ud-*.conllu".format(s,d))
+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[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))}
+  __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)
+lid=trainDS.label2id
+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,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)
+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 sha256:990574f354bc5c07915f67c322132900a90fa675991ebc32d1de2c39f03a34a8
+size 351720561

sentencepiece.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:01ba4719c80b6fe911b091a7c05124b64eeece964e09c058ef8f9805daca546b
+size 1

special_tokens_map.json

@@ -0,0 +1,15 @@
+{
+  "bos_token": "[CLS]",
+  "cls_token": "[CLS]",
+  "eos_token": "[SEP]",
+  "mask_token": {
+    "content": "[MASK]",
+    "lstrip": true,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": "[UNK]"
+}

tokenizer_config.json

@@ -0,0 +1,21 @@
+{
+  "bos_token": "[CLS]",
+  "cls_token": "[CLS]",
+  "do_lower_case": false,
+  "eos_token": "[SEP]",
+  "keep_accents": true,
+  "mask_token": {
+    "__type": "AddedToken",
+    "content": "[MASK]",
+    "lstrip": true,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "model_max_length": 512,
+  "pad_token": "[PAD]",
+  "remove_space": true,
+  "sep_token": "[SEP]",
+  "tokenizer_class": "RemBertTokenizerFast",
+  "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