请问如何使用hf加载icetk_glm_130B的tokenizer和GLM130B的模型?
Ajay-Wong opened this issue · comments
As I know, there is no huggingface support for GLM130B. Enjoy using it in SAT :)
If there is any question, feel free to post here and I will fix it as soon as possible.
As I know, there is no huggingface support for GLM130B. Enjoy using it in SAT :)
If there is any question, feel free to post here and I will fix it as soon as possible.
好的,谢谢,那再问下,chatglm6b与130b模型结构上除了多一些层和维度的差别?是不是可以使用chatglm6b的方式改下模型参数,直接加载130b的模型?
具体使用方法可以参考GLM130B的仓库:https://github.com/THUDM/GLM-130B
具体使用方法可以参考GLM130B的仓库:https://github.com/THUDM/GLM-130B
谢谢,这个看到了,有没有什么资料是介绍130b与chatglm6b或者chatglm2-6b的有什么区别的?
模型代码都在这里,可以自行比较:https://github.com/THUDM/SwissArmyTransformer/tree/main/sat/model/official
每个模型代码都是100行左右,很容易对比。
模型代码都在这里,可以自行比较:https://github.com/THUDM/SwissArmyTransformer/tree/main/sat/model/official
每个模型代码都是100行左右,很容易对比。
thanks