Prerequisites
Feature Description
The ability to continue generation after a stop.
which can
Recovering from accidental interruptions , and
Editing or rewriting hallucinated output, then continuing the reasoning
Please consider adding a "Continue" button in the WebUI that allows users to resume generation from the last stopped point.
Motivation
editing or rewriting hallucinated output, then continuing the reasoning seamlessly
and
recovering from accidental interruptions without restarting the entire prompt
Currently, I must either manually reconstruct the prompt or re-inject the previous output to resume generation. For models that rely on multi-step CoT or thinking model, this approach consumes a significant number of tokens and may degrade performance. A native "Continue Generation" feature would preserve context more efficiently.(which are very slow my pc only have 3 token/s )
Possible Implementation
No response . But i see LM studio can do it
Prerequisites
Feature Description
The ability to continue generation after a stop.
which can
Recovering from accidental interruptions , and
Editing or rewriting hallucinated output, then continuing the reasoning
Please consider adding a "Continue" button in the WebUI that allows users to resume generation from the last stopped point.
Motivation
editing or rewriting hallucinated output, then continuing the reasoning seamlessly
and
recovering from accidental interruptions without restarting the entire prompt
Currently, I must either manually reconstruct the prompt or re-inject the previous output to resume generation. For models that rely on multi-step CoT or thinking model, this approach consumes a significant number of tokens and may degrade performance. A native "Continue Generation" feature would preserve context more efficiently.(which are very slow my pc only have 3 token/s )
Possible Implementation
No response . But i see LM studio can do it