ai-forever / ai-forever/Kandinsky-2

tokenizer for 2.2 is not mclip but clip?

未关闭
#106 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看
主要语言
Jupyter Notebook
星标
2.8k
派生
319
PR 合并指标
30 天内没有已合并 PR

描述

Hi, I tried to generate images with Russian, Chinese, and Japanese input text, but Kandinsky 2.2 doesn't seem to understand these languages. So I looked into the code and tried to debug and noticed that the tokenizer is actually clip not mclip. Could you please look into this?

贡献指南

这个仓库没有索引到贡献指南

评估

这个 Issue 还没有评估数据。

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。