Closed IzzyPutterman closed 1 month ago
@IzzyPutterman was this breaking the ITL calculation (e.g. empty output --> zero token --> divide by zero)? Just trying to understand the context.
@IzzyPutterman was this breaking the ITL calculation (e.g. empty output --> zero token --> divide by zero)? Just trying to understand the context.
(github not letting me reply), but empty-output -> zero token -> replaced by a 1 token instead in llm_metrics-> total number of tokens used for ITL calculation is > that total expected, but thats not reported. So ITL is smaller than expected.
@IzzyPutterman was this breaking the ITL calculation (e.g. empty output --> zero token --> divide by zero)? Just trying to understand the context.
(github not letting me reply), but empty-output -> zero token -> replaced by a 1 token instead in llm_metrics-> total number of tokens used for ITL calculation is > that total expected, but thats not reported. So ITL is smaller than expected.
Ah I see. Yeah that makes sense. I think these will be irrelevant once we update to a new ITL formula, but the change looks good to me.
In vLLM outputs can be empty in the middle of the stream.