您好,我刚接触深度学习不久,所以问题比较简单,希望您别介意。
问题一:
我遵循了您readme中的步骤,现在正在运行main.py,但是一直卡在epoch 0 loss 就不往下进行了,运行过程如下所示
---Regenerate Data---
49
{0: 'UNK', 1: '主演', 2: '歌手', 3: '简称', 4: '总部地点', 5: '导演', 6: '出生地', 7: '目', 8: '出生日期', 9: '占地面积', 10: '上映时间', 11: '出版社', 12: '作者', 13: '号', 14: '父亲', 15: '毕业院校', 16: '成立日期', 17: '改编自', 18: '主持人', 19: '所属专辑', 20: '连载网站', 21: '作词', 22: '作曲', 23: '创始人', 24: '丈夫', 25: '妻子', 26: '朝代', 27: '民族', 28: '国籍', 29: '身高', 30: '出品公司', 31: '母亲', 32: '编剧', 33: '首都', 34: '面积', 35: '祖籍', 36: '嘉宾', 37: '字', 38: '海拔', 39: '注册资本', 40: '制片人', 41: '董事长', 42: '所在城市', 43: '气候', 44: '人口数量', 45: '邮政编码', 46: '主角', 47: '官方语言', 48: '修业年限'}
Some weights of the model checkpoint at ./bert-base-chinese were not used when initializing BertModel: ['cls.predictions.decoder.weight', 'cls.seq_relationship.bias', 'cls.seq_relationship.weight', 'cls.predictions.transform.dense.weight', 'cls.predictions.transform.LayerNorm.bias', 'cls.predictions.transform.dense.bias', 'cls.predictions.bias', 'cls.predictions.transform.LayerNorm.weight']
- This IS expected if you are initializing BertModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing BertModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
--train data--
torch.Size([200, 1, 128])
torch.Size([200, 1, 128])
torch.Size([200])
--train data--
torch.Size([200, 1, 128])
torch.Size([200, 1, 128])
torch.Size([200])
--eval data--
torch.Size([45558, 1, 128])
torch.Size([45558, 1, 128])
torch.Size([45558])
epoch 0 loss: 2.12479829788208 right 51 total 200 Acc: 0.255
问题二:
另外,请问一下,看了您的readme,如果能够跑通,或者直接使用您的模型,修改test,然后运行demo,就可以了吧?