def __init__(self, dataset):
self.model_name = 'bert'
# self.train_path = dataset + '/data/train.txt' # 训练集
# self.dev_path = dataset + '/data/dev.txt' # 验证集
# self.test_path = dataset + '/data/test.txt' # 测试集
self.train_path = dataset + '/data/train.bin' # 训练集
self.dev_path = dataset + '/data/dev.bin' # 验证集
self.test_path = dataset + '/data/test.bin' # 测试集
self.class_list = [
x.strip() for x in open(dataset + '/data/class.txt').readlines()
] # 类别名单
self.save_path = dataset + '/saved_dict/' + self.model_name + '.ckpt' # 模型训练结果
self.device = torch.device(
'cuda' if torch.cuda.is_available() else 'cpu') # 设备
self.require_improvement = 1000 # 若超过1000batch效果还没提升，则提前结束训练
self.num_classes = len(self.class_list) # 类别数
self.num_epochs = 3 # epoch数
self.batch_size = 8 # mini-batch大小
self.pad_size = 100 # 每句话处理成的长度(短填长切)
self.learning_rate = 1e-5 # 学习率
self.adam_epsilon = 1e-6 # eps
self.bert_path = './albert_tiny'
# self.tokenizer = BertTokenizer.from_pretrained(self.bert_path)
self.tokenizer = AlbertTokenizer.from_pretrained(
'./albert_tiny/vocab.txt')
# self.hidden_size = 768
self.hidden_size = 312
def __init__(self):
self.save_path = basedir + '/models/albert.ckpt' # 模型训练结果
self.class_list = [
x.strip() for x in open(basedir + '/models/class.txt').readlines()
] # 类别名单
self.device = torch.device(
'cuda' if torch.cuda.is_available() else 'cpu') # 设备
self.num_classes = len(self.class_list) # 类别数
self.bert_path = './albert_tiny'
self.tokenizer = AlbertTokenizer.from_pretrained(self.bert_path)
self.hidden_size = 312
self.pad_size = 100 # 每句话处理成的长度(短填长切)
# self.learning_rate = 1e-5 # 学习率
# self.adam_epsilon = 1e-6 # 学习率
self.PAD, self.CLS = '[PAD]', '[CLS]'
from albert_zh import AlbertConfig,AlbertForSequenceClassification,AlbertTokenizer
import torch
if __name__ == "__main__":
tokenizer = AlbertTokenizer.from_pretrained('albert_tiny/vocab.txt')
model_config = AlbertConfig.from_json_file('./albert_tiny/config.json')
model = AlbertForSequenceClassification.from_pretrained('./albert_tiny',config = model_config)
intput_str = '周杰倫，臺灣著名華語流行歌曲男歌手、音樂家、唱片製片人。同時是演員、導演，也是電競團隊隊長兼老闆、服飾品牌老闆。以其個人風格和聲歌手樂創作能力著稱，影響華語樂壇。 在2000年，周杰倫發行了他的首張專輯《Jay》，從屬於唱片公司阿爾發音樂。'
input_ids = torch.tensor(tokenizer.encode(intput_str)).unsqueeze(0) # Batch size 1
labels = torch.tensor([1]).unsqueeze(0) # Batch size 1
outputs = model(input_ids, labels=labels)
loss, logits = outputs[:2]
print(loss,logits)