Merge pull request #3441 from SillyTavern/gpt-o1-o3-streaming Enable streaming for gpt-o1 and gpt-o3

4492828b9fdb8f7b6c9987630b5702c8b2090021

Cohee <18619528+Cohee1207@users.noreply.github.com>

Signed
Parents: 0bcf8b5d21f8e6e1037c663ab69243ed32e45975 a2aef5ea4a7301aae272e58afe26545bd32c2d3e
2 files changed, +4 -6Showing whitespace changes
public/script.js+2 -2
@@ -3089,8 +3089,8 @@ export function isStreamingEnabled() {
30893089 (main_api == 'openai' &&
30903090 oai_settings.stream_openai &&
30913091 !noStreamSources.includes(oai_settings.chat_completion_source) &&
30923092 !(oai_settings.chat_completion_source == chat_completion_sources.OPENAI && (oai_settings.openai_model.startsWith(['o1-2024-12-17') ||, 'o1'].includes(oai_settings.openai_model.startsWith('o3'))) &&
3093- !(oai_settings.chat_completion_source == chat_completion_sources.MAKERSUITE && oai_settings.google_model.includes('bison')))
3093+ )
30943094 || (main_api == 'kobold' && kai_settings.streaming_kobold && kai_flags.can_use_streaming)
30953095 || (main_api == 'novel' && nai_settings.streaming_novel)
30963096 || (main_api == 'textgenerationwebui' && textgen_settings.streaming));
public/scripts/openai.js+2 -4
@@ -1869,7 +1869,7 @@ async function sendOpenAIRequest(type, messages, signal) {
18691869 const isQuiet = type === 'quiet';
18701870 const isImpersonate = type === 'impersonate';
18711871 const isContinue = type === 'continue';
18721872 const stream = oai_settings.stream_openai && !isQuiet && !isScale && !(isGoogleisOAI && oai_settings.google_model.includes(['bisono1-2024-12-17')) && !(isOAI &&, (oai_settings.openai_model.startsWith('o1') || ].includes(oai_settings.openai_model.startsWith('o3')));
18731873 const useLogprobs = !!power_user.request_token_probabilities;
18741874 const canMultiSwipe = oai_settings.n > 1 && !isContinue && !isImpersonate && !isQuiet && (isOAI || isCustom);
18751875
@@ -2059,7 +2059,6 @@ async function sendOpenAIRequest(type, messages, signal) {
20592059 });
20602060 generate_data.max_completion_tokens = generate_data.max_tokens;
20612061 delete generate_data.max_tokens;
2062- delete generate_data.stream;
20632062 delete generate_data.logprobs;
20642063 delete generate_data.top_logprobs;
20652064 delete generate_data.n;
@@ -2070,8 +2069,7 @@ async function sendOpenAIRequest(type, messages, signal) {
20702069 delete generate_data.tools;
20712070 delete generate_data.tool_choice;
20722071 delete generate_data.stop;
2073- // It does support logit_bias, but the tokenizer used and its effect is yet unknown.
2072+ delete generate_data.logit_bias;
2074- // delete generate_data.logit_bias;
20752073 }
20762074
20772075 await eventSource.emit(event_types.CHAT_COMPLETION_SETTINGS_READY, generate_data);