Adjust number of VLLM logprobs
This commit is contained in:
parent
60b7164c28
commit
1dd21caa66
|
@ -1019,6 +1019,14 @@ export function isJsonSchemaSupported() {
|
||||||
return [TABBY, LLAMACPP].includes(settings.type) && main_api === 'textgenerationwebui';
|
return [TABBY, LLAMACPP].includes(settings.type) && main_api === 'textgenerationwebui';
|
||||||
}
|
}
|
||||||
|
|
||||||
|
function getLogprobsNumber() {
|
||||||
|
if (settings.type === VLLM) {
|
||||||
|
return 5;
|
||||||
|
}
|
||||||
|
|
||||||
|
return 10;
|
||||||
|
}
|
||||||
|
|
||||||
export function getTextGenGenerationData(finalPrompt, maxTokens, isImpersonate, isContinue, cfgValues, type) {
|
export function getTextGenGenerationData(finalPrompt, maxTokens, isImpersonate, isContinue, cfgValues, type) {
|
||||||
const canMultiSwipe = !isContinue && !isImpersonate && type !== 'quiet';
|
const canMultiSwipe = !isContinue && !isImpersonate && type !== 'quiet';
|
||||||
const { banned_tokens, banned_strings } = getCustomTokenBans();
|
const { banned_tokens, banned_strings } = getCustomTokenBans();
|
||||||
|
@ -1028,7 +1036,7 @@ export function getTextGenGenerationData(finalPrompt, maxTokens, isImpersonate,
|
||||||
'model': getTextGenModel(),
|
'model': getTextGenModel(),
|
||||||
'max_new_tokens': maxTokens,
|
'max_new_tokens': maxTokens,
|
||||||
'max_tokens': maxTokens,
|
'max_tokens': maxTokens,
|
||||||
'logprobs': power_user.request_token_probabilities ? 10 : undefined,
|
'logprobs': power_user.request_token_probabilities ? getLogprobsNumber() : undefined,
|
||||||
'temperature': settings.dynatemp ? (settings.min_temp + settings.max_temp) / 2 : settings.temp,
|
'temperature': settings.dynatemp ? (settings.min_temp + settings.max_temp) / 2 : settings.temp,
|
||||||
'top_p': settings.top_p,
|
'top_p': settings.top_p,
|
||||||
'typical_p': settings.typical_p,
|
'typical_p': settings.typical_p,
|
||||||
|
|
Loading…
Reference in New Issue