initial release

README.md

@@ -0,0 +1,33 @@
+---
+language:
+- "th"
+tags:
+- "thai"
+- "token-classification"
+- "pos"
+- "wikipedia"
+license: "apache-2.0"
+pipeline_tag: "token-classification"
+widget:
+- text: "หลายหัวดีกว่าหัวเดียว"
+---
+
+# roberta-base-thai-syllable-upos
+
+## Model Description
+
+This is a RoBERTa model pre-trained on Thai Wikipedia texts for POS-tagging, derived from [roberta-base-thai-syllable](https://huggingface.co/KoichiYasuoka/roberta-base-thai-syllable). Every word is tagged by [UPOS](https://universaldependencies.org/u/pos/) (Universal Part-Of-Speech).
+
+## How to Use
+
+```py
+import torch
+from transformers import AutoTokenizer,AutoModelForTokenClassification
+tokenizer=AutoTokenizer.from_pretrained("KoichiYasuoka/roberta-base-thai-syllable-upos")
+model=AutoModelForTokenClassification.from_pretrained("KoichiYasuoka/roberta-base-thai-syllable-upos")
+s="หลายหัวดีกว่าหัวเดียว"
+t=tokenizer.tokenize(s)
+p=[model.config.id2label[q] for q in torch.argmax(model(tokenizer.encode(s,return_tensors="pt"))[0],dim=2)[0].tolist()[1:-1]]
+print(list(zip(t,p)))
+```
+

config.json

@@ -0,0 +1,116 @@
+{
+  "architectures": [
+    "RobertaForTokenClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": null,
+  "eos_token_id": 2,
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "id2label": {
+    "0": "B-NUM",
+    "1": "I-NUM",
+    "2": "PART",
+    "3": "B-NOUN",
+    "4": "B-DET",
+    "5": "PROPN",
+    "6": "ADJ",
+    "7": "I-ADP",
+    "8": "B-VERB",
+    "9": "I-ADV",
+    "10": "PRON",
+    "11": "I-AUX",
+    "12": "I-PROPN",
+    "13": "B-CCONJ",
+    "14": "DET",
+    "15": "B-ADP",
+    "16": "SYM",
+    "17": "I-DET",
+    "18": "B-AUX",
+    "19": "AUX",
+    "20": "VERB",
+    "21": "I-VERB",
+    "22": "I-PART",
+    "23": "B-PROPN",
+    "24": "I-PUNCT",
+    "25": "I-PRON",
+    "26": "CCONJ",
+    "27": "ADV",
+    "28": "NUM",
+    "29": "SCONJ",
+    "30": "PUNCT",
+    "31": "NOUN",
+    "32": "I-ADJ",
+    "33": "B-PRON",
+    "34": "B-ADJ",
+    "35": "B-PART",
+    "36": "B-ADV",
+    "37": "I-NOUN",
+    "38": "ADP",
+    "39": "B-PUNCT",
+    "40": "I-CCONJ"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "label2id": {
+    "ADJ": 6,
+    "ADP": 38,
+    "ADV": 27,
+    "AUX": 19,
+    "B-ADJ": 34,
+    "B-ADP": 15,
+    "B-ADV": 36,
+    "B-AUX": 18,
+    "B-CCONJ": 13,
+    "B-DET": 4,
+    "B-NOUN": 3,
+    "B-NUM": 0,
+    "B-PART": 35,
+    "B-PRON": 33,
+    "B-PROPN": 23,
+    "B-PUNCT": 39,
+    "B-VERB": 8,
+    "CCONJ": 26,
+    "DET": 14,
+    "I-ADJ": 32,
+    "I-ADP": 7,
+    "I-ADV": 9,
+    "I-AUX": 11,
+    "I-CCONJ": 40,
+    "I-DET": 17,
+    "I-NOUN": 37,
+    "I-NUM": 1,
+    "I-PART": 22,
+    "I-PRON": 25,
+    "I-PROPN": 12,
+    "I-PUNCT": 24,
+    "I-VERB": 21,
+    "NOUN": 31,
+    "NUM": 28,
+    "PART": 2,
+    "PRON": 10,
+    "PROPN": 5,
+    "PUNCT": 30,
+    "SCONJ": 29,
+    "SYM": 16,
+    "VERB": 20
+  },
+  "layer_norm_eps": 1e-05,
+  "mask_token_id": 4,
+  "max_position_embeddings": 514,
+  "model_type": "roberta",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 1,
+  "position_embedding_type": "absolute",
+  "tokenizer_class": "BertTokenizerFast",
+  "torch_dtype": "float32",
+  "transformers_version": "4.10.0",
+  "type_vocab_size": 1,
+  "unk_token_id": 3,
+  "use_cache": true,
+  "vocab_size": 33144
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4860ad4ef9b423ca8a89f032a9b835d23ecbc82ae21dac6e9a6c8279241959c
+size 443838354

special_tokens_map.json

@@ -0,0 +1 @@
+{"unk_token": "<unk>", "sep_token": "</s>", "pad_token": "<pad>", "cls_token": "<s>", "mask_token": "<mask>"}

tokenizer_config.json

@@ -0,0 +1 @@
+{"do_lower_case": false, "unk_token": "<unk>", "sep_token": "</s>", "pad_token": "<pad>", "cls_token": "<s>", "mask_token": "<mask>", "tokenize_chinese_chars": true, "strip_accents": false, "model_max_length": 416, "do_basic_tokenize": true, "never_split": null, "tokenizer_class": "BertTokenizer"}