LISA
LISA copied to clipboard
[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?