如何更好的微调
#12175
Replies: 3 comments
-
这种情况不需要覆盖字典中的所有字符,有针对性的补充罗马字母的数据即可,5万左右的数据,迭代轮次少一点。最好构建一个比较准确的评估集,防止模型学偏。 |
Beta Was this translation helpful? Give feedback.
0 replies
-
评估集要包含所有字符吗?需要冻结前面的层吗?谢谢 |
Beta Was this translation helpful? Give feedback.
0 replies
-
请问你是如何处理字典的,有新增需要识别的字符,是直接加在原来的模型字典吗,字典长度改变就不能加载预训练参数了 |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
请问大神们,使用中英文通用模型的预训练模型进行预测,对于特定字符识别识别效果较差(比如罗马字母),我想模型进行微调(不改变字典),需要冻结前面的层,只对最后FC层进行训练吗?微调的数据集需要覆盖字典的所有字符吗?每个字符的数据个数多少比较合适,谢谢
Beta Was this translation helpful? Give feedback.
All reactions