[ehealth] fix metric and ehealth name

Author: LemonNoel

examples/biomedical/cblue/README.md

@@ -10,7 +10,7 @@
 
 ## 数据集介绍
 
-本项目使用了中文医学语言理解测评（[Chinese Biomedical Language Understanding Evaluation，CBLUE](https://github.com/CBLUEbenchmark/CBLUE)）数据集，[<sup>[1]</sup>](#refer-anchor-cblue)其包括医学文本信息抽取（实体识别、关系抽取）、医学术语归一化、医学文本分类、医学句子关系判定和医学问答共5大类任务8个子任务。
+本项目使用了中文医学语言理解测评（[Chinese Biomedical Language Understanding Evaluation，CBLUE](https://github.com/CBLUEbenchmark/CBLUE)）数据集，其包括医学文本信息抽取（实体识别、关系抽取）、医学术语归一化、医学文本分类、医学句子关系判定和医学问答共5大类任务8个子任务。
 
 * CMeEE：中文医学命名实体识别
 * CMeIE：中文医学文本实体关系抽取
@@ -81,5 +81,4 @@ $ python -m paddle.distributed.launch --gpus "0,1,2,3" train.py --dataset CHIP-C
 pip install xlrd==1.2.0
 ```
 
-<div id="refer-anchor-cblue"></div>
-- [1] CBLUE: A Chinese Biomedical Language Understanding Evaluation Benchmark [pdf](https://arxiv.org/abs/2106.08087) [git](https://github.com/CBLUEbenchmark/CBLUE) [web](https://tianchi.aliyun.com/specials/promotion/2021chinesemedicalnlpleaderboardchallenge)
+[1] CBLUE: A Chinese Biomedical Language Understanding Evaluation Benchmark [pdf](https://arxiv.org/abs/2106.08087) [git](https://github.com/CBLUEbenchmark/CBLUE) [web](https://tianchi.aliyun.com/specials/promotion/2021chinesemedicalnlpleaderboardchallenge)

examples/biomedical/cblue/train_classification.py

@@ -26,7 +26,7 @@
 import paddlenlp as ppnlp
 from paddlenlp.data import Stack, Tuple, Pad
 from paddlenlp.datasets import load_dataset
-from paddlenlp.transformers import LinearDecayWithWarmup
+from paddlenlp.transformers import ElectraForSequenceClassification, LinearDecayWithWarmup
 from paddlenlp.metrics import MultiLabelsMetric, AccuracyAndF1
 from paddlenlp.ops.optimizer import ExponentialMovingAverage
 
@@ -143,13 +143,13 @@ def do_train():
     train_ds, dev_ds, test_ds = load_dataset(
         'cblue', args.dataset, splits=['train', 'dev', 'test'])
 
-    model = ppnlp.transformers.ElectraForSequenceClassification.from_pretrained(
-        'chinese-ehealth',
+    model = ElectraForSequenceClassification.from_pretrained(
+        'ehealth-chinese',
         num_classes=len(train_ds.label_list),
         activation='tanh',
         layer_norm_eps=1e-5)
     tokenizer = ppnlp.transformers.ElectraTokenizer.from_pretrained(
-        'chinese-ehealth')
+        'ehealth-chinese')
 
     trans_func = partial(
         convert_example,

paddlenlp/transformers/electra/modeling.py

@@ -226,7 +226,7 @@ class ElectraPretrainedModel(PretrainedModel):
             "type_vocab_size": 2,
             "vocab_size": 21128
         },
-        "chinese-ehealth": {
+        "ehealth-chinese": {
             "attention_probs_dropout_prob": 0.1,
             "embedding_size": 768,
             "hidden_act": "gelu",
@@ -255,8 +255,8 @@ class ElectraPretrainedModel(PretrainedModel):
             "https://bj.bcebos.com/paddlenlp/models/transformers/chinese-electra-small/chinese-electra-small.pdparams",
             "chinese-electra-base":
             "https://bj.bcebos.com/paddlenlp/models/transformers/chinese-electra-base/chinese-electra-base.pdparams",
-            "chinese-ehealth":
-            "https://paddlenlp.bj.bcebos.com/models/transformers/chinese-ehealth.pdparams"
+            "ehealth-chinese":
+            "https://paddlenlp.bj.bcebos.com/models/transformers/ehealth_chinese/ehealth-chinese.pdparams"
         }
     }
 

paddlenlp/transformers/electra/tokenizer.py

@@ -80,8 +80,8 @@ class ElectraTokenizer(PretrainedTokenizer):
             "http://bj.bcebos.com/paddlenlp/models/transformers/chinese-electra-base/vocab.txt",
             "chinese-electra-small":
             "http://bj.bcebos.com/paddlenlp/models/transformers/chinese-electra-small/vocab.txt",
-            "chinese-ehealth":
-            "https://paddlenlp.bj.bcebos.com/models/transformers/chinese_ehealth/vocab.txt"
+            "ehealth-chinese":
+            "https://paddlenlp.bj.bcebos.com/models/transformers/ehealth_chinese/vocab.txt"
         }
     }
     pretrained_init_configuration = {
@@ -100,7 +100,7 @@ class ElectraTokenizer(PretrainedTokenizer):
         "chinese-electra-small": {
             "do_lower_case": True
         },
-        "chinese-ehealth": {
+        "ehealth-chinese": {
             "do_lower_case": True
         }
     }