xiexing0916

Results 1 issues of xiexing0916

[seg] token inference

I see that the [seg] token in the inference code is given directly in the conversation prompt, and not from next token prediction?