bebilli
bebilli
### Self Checks - [x] This is only for bug report, if you would like to ask a question, please head to [Discussions](https://github.com/langgenius/dify/discussions/categories/general). - [x] I have searched for existing...
### Self Checks - [x] This is only for bug report, if you would like to ask a question, please head to [Discussions](https://github.com/langgenius/dify/discussions/categories/general). - [x] I have searched for existing...
In the example of "talk_to_claude", on the interface, each response from the AI consists of two repeated messages, and there are no chat records of the user. I've looked at...
### Self Checks - [x] This is only for bug report, if you would like to ask a question, please head to [Discussions](https://github.com/langgenius/dify/discussions/categories/general). - [x] I have searched for existing...
Notice: In order to resolve issues more efficiently, please raise issue following the template. （注意：为了更加高效率解决您遇到的问题，请按照模板提问，补充细节） ## ❓ Questions and Help 我查看过往的issue，发现model.generate 不支持异步或多线程，这样的话就需要加锁，或为每一个任务分别加载一次model，在高并发下，这样会显著降低吞吐量。如果因为GIL 问题没有意义，是否在model对象的内部设计一个进程池，来提高吞吐量，而不是交给用户去多进程加载N次模型从而导致显存无意义的冗余占用？ #### What's your environment? - OS (Windows)...
Run in GPU mode (4090), with the embedded model being google/embeddinggemma-300m; use the built-in test cases for testing. 1、The built-in test cases fail to recall the problem. Please refer to...