Use new inference model config / API #2208
Conversation
| parent_id, | ||
| work_parameters: { | ||
| ...params, | ||
| top_k: params.top_k === null ? null : Math.floor(Math.pow(32000, params.top_k)), |
There was a problem hiding this comment.
Have you tested the top_k here? Andreas suggest an exponential function for this.
There was a problem hiding this comment.
ah nvm I see the config below
There was a problem hiding this comment.
Yeah just for info: this works for the custom config where the slider goes between 0 and 1, but for all the presets it computes pow(32000, 50) or similar.
I did not find a clean way of changing this so I changed min and max. Arguably, no more than 1000 is needed, because the sampling probability is already so low it might as well be zero.
| const form = useForm<WorkParametersInput>({ | ||
| defaultValues: modelInfos[0].parameter_configs[0].work_parameters, | ||
|
|
||
| console.assert(modelInfos.length > 0, "No model config was found"); |
There was a problem hiding this comment.
I think we can ignore this and assume the backend always returns at least 1 config.
There was a problem hiding this comment.
Yes, this console.assert is there to inform other developers that we expect the array to have at least one. You and I know that, any new person who might look over this might not know this.
AbdBarho
force-pushed
the
use-model-config
branch
from
0739957
to
4ad687b
Compare
Blocked by #2207