Field 'backend_sampling': [json.exception.type_error.302] type must be boolean, but is string
what happened!?
Could you please provide more details about the circumstances under which the error occurred?
This error is caused by a string value (such as "true" or "false" enclosed in quotes) being specified for the backend_sampling parameter in the llama.cpp configuration—typically within a configuration file like config.json or an API JSON request—where a boolean value (true or false) should have been used instead.
Could you please provide more details about the circumstances under which the error occurred?
This error is caused by a string value (such as "true" or "false" enclosed in quotes) being specified for thebackend_samplingparameter in the llama.cpp configuration—typically within a configuration file likeconfig.jsonor an API JSON request—where a boolean value (trueorfalse) should have been used instead.
Thank you for providing detailed logs.
The primary focus of the llama.cpp developers is on optimizing the limits of the C++ inference engine. The Web UI is treated merely as an auxiliary tool for functionality verification and is rapidly developed and patched by the community.
As a result, occasional basic front-end bugs may slip in—such as the current backend_sampling issue where JSON data types are mismatched (sending a string instead of a boolean), or certain parameters failing to be correctly passed from the UI to the server.
I have rebuilt the Space using the "Factory rebuild" option to update llama.cpp to the latest version, which should resolve the issue. I also recommend performing a hard refresh of your browser. Additionally, I have adjusted the context length to 32,000 tokens to further improve inference speed.
