朱嘉伟

python demo.py zsh: segmentation fault

2

sorry i'm not a native english speaker, so let me describe the problem simply: git clone --recursive https://github.com/andrewssobral/bgslibrary.git bgs cd bgs/build python setup.py build > ... [100%] Built target bgs_python...

似乎是有赞的变动造成的 我不知道该怎么解决

npm run start > [email protected] start C:\Users\Jiawei\Desktop\YouzanPayPortal-master > mkdirp logs && npm run build && cross-env NODE_ENV=production forever start dist/app.js > [email protected] build C:\Users\Jiawei\Desktop\YouzanPayPortal-master > rimraf dist && tsc warn:...

Unable to create interpreter

5

``` I/tflite (10801): Created TensorFlow Lite delegate for GPU. I/tflite (10801): Initialized TensorFlow Lite runtime. E/tflite (10801): Following operations are not supported by GPU delegate: E/tflite (10801): EXP: Operation is...

能否解决下这个问题，顺便兼容下手机浏览器？ 请问是用的什么模型？是用onnx构建的嘛？我想自己试着调试下看能不能解决

It seems that AnyText only supports max image size: 512*512?

An error occurred during the Push To Replicate phase: ``` #... #14 8.080 Downloading websockets-12.0-cp39-cp39-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_17_x86_64.manylinux2014_x86_64.whl (130 kB) #14 8.354 Installing collected packages: typing-extensions, sniffio, idna, exceptiongroup, h11, click, anyio, websockets,...

请问train_cnt多少合适？

1

代码里的是 train_cnt = 100 val_cnt = 5 test_cnt = 30 然后生成的每个字体的对应训练图片是100个，但是中文三千多个汉字的话100张图片能覆盖到每个字吗？

请问是否支持基于前一次训练的ckpt再次训练新增的字体？

2

大佬你好 我第一次在config里设置FONT_COUNT是1000，现在想改为2000因为新增了1000个字体，并且想节约成本直接复用前一次的训练成果，但是这一步似乎有问题呀，输出维度和字体索引映射都固化了，岂不是不支持二次训练？ 有方法可以实现吗？