Replies: 3 comments 9 replies
-
试试main分支,2.8版本可能有训练精度的问题。 |
Beta Was this translation helpful? Give feedback.
1 reply
-
或者试试在2.7指定一下numpy版本 |
Beta Was this translation helpful? Give feedback.
5 replies
-
我看第一张图,使用的是C盘里python环境的numpy,这可能是导致numpy版本报错的原因 |
Beta Was this translation helpful? Give feedback.
3 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.
-
AttributeError:

np.sctypes
was removed in the NumPy 2.0 release. Access dtypes explicitly instead.2.7.0版本训练效果还不错,acc可以到达0.9,python端推理顺利,但是导出模型一直会报错numpy版本问题。实际安装numpy<2.0。
2.7.0可以训练,无法导出。
2.7.0使用的预训练模型:en_PP-OCRv3_rec_train
2.8.0训练和导出都可以实现,但是训练过程中acc一直为0。

2.8.0使用的预训练模型:en_PP-OCRv4_rec_train
2.7.0和2.8.0数据集一样,2.7.0训练顺利,无法导出。2.8.0训练acc始终为0,会保存best_model且可以顺利导出。
2.7.0训练速度>2.8.0训练速度。
Beta Was this translation helpful? Give feedback.
All reactions