Gemma 3 (#3686) * Gemma 3 * Adjust safetySettings * Disable sysprompt * Add isGemma check to tool processing logic * Disable a google search tool for gemma
Signed| @@ -3134,6 +3134,9 @@ | ||
| 3134 | 3134 | <option value="gemini-ultra">Gemini Ultra (1.0)</option> |
| 3135 | 3135 | <option value="gemini-1.0-ultra-latest">Gemini 1.0 Ultra</option> |
| 3136 | 3136 | </optgroup> |
| 3137 | + <optgroup label="Gemma"> | |
| 3138 | + <option value="gemma-3-27b-it">Gemma 3 27B</option> | |
| 3139 | + </optgroup> | |
| 3137 | 3140 | <optgroup label="Subversions"> |
| 3138 | 3141 | <option value="gemini-2.0-pro-exp">Gemini 2.0 Pro Experimental</option> |
| 3139 | 3142 | <option value="gemini-2.0-pro-exp-02-05">Gemini 2.0 Pro Experimental 2025-02-05</option> |
| @@ -4341,10 +4341,12 @@ async function onModelChange() { | ||
| 4341 | 4341 | $('#openai_max_context').attr('max', max_32k); |
| 4342 | 4342 | } else if (value.includes('gemini-1.0-ultra') || value === 'gemini-ultra') { |
| 4343 | 4343 | $('#openai_max_context').attr('max', max_32k); |
| 4344 | + } else if (value.includes('gemma-3')) { | |
| 4345 | + $('#openai_max_context').attr('max', max_128k); | |
| 4344 | 4346 | } else { |
| 4345 | 4347 | $('#openai_max_context').attr('max', max_4k); |
| 4346 | 4348 | } |
| 4347 | 4349 | let makersuite_max_temp = (value.includes('vision') || value.includes('ultra') || value.includes('gemma')) ? 1.0 : 2.0; |
| 4348 | 4350 | oai_settings.temp_openai = Math.min(makersuite_max_temp, oai_settings.temp_openai); |
| 4349 | 4351 | $('#temp_openai').attr('max', makersuite_max_temp).val(oai_settings.temp_openai).trigger('input'); |
| 4350 | 4352 | oai_settings.openai_max_context = Math.min(Number($('#openai_max_context').attr('max')), oai_settings.openai_max_context); |
| @@ -340,6 +340,7 @@ async function sendMakerSuiteRequest(request, response) { | ||
| 340 | 340 | const enableWebSearch = Boolean(request.body.enable_web_search); |
| 341 | 341 | const requestImages = Boolean(request.body.request_images); |
| 342 | 342 | const isThinking = model.includes('thinking'); |
| 343 | + const isGemma = model.includes('gemma'); | |
| 343 | 344 | |
| 344 | 345 | const generationConfig = { |
| 345 | 346 | stopSequences: request.body.stop, |
| @@ -375,8 +376,8 @@ async function sendMakerSuiteRequest(request, response) { | ||
| 375 | 376 | const prompt = convertGooglePrompt(request.body.messages, model, useSystemPrompt, getPromptNames(request)); |
| 376 | 377 | let safetySettings = GEMINI_SAFETY; |
| 377 | 378 | |
| 378 | 379 | // These old models do not support setting the threshold to OFF at all. |
| 379 | 380 | if (['gemini-1.5-pro-001', 'gemini-1.5-flash-001', 'gemini-1.5-flash-8b-exp-0827', 'gemini-1.5-flash-8b-exp-0924', 'gemini-pro', 'gemini-1.0-pro', 'gemini-1.0-pro-001', 'gemma-3-27b-it'].includes(model)) { |
| 380 | 381 | safetySettings = GEMINI_SAFETY.map(setting => ({ ...setting, threshold: 'BLOCK_NONE' })); |
| 381 | 382 | } |
| 382 | 383 | // Interestingly, Gemini 2.0 Flash does support setting the threshold for HARM_CATEGORY_CIVIC_INTEGRITY to OFF. |
| @@ -385,14 +386,14 @@ async function sendMakerSuiteRequest(request, response) { | ||
| 385 | 386 | } |
| 386 | 387 | // Most of the other models allow for setting the threshold of filters, except for HARM_CATEGORY_CIVIC_INTEGRITY, to OFF. |
| 387 | 388 | |
| 388 | 389 | if (enableWebSearch && !useMultiModal && !isGemma) { |
| 389 | 390 | const searchTool = model.includes('1.5') || model.includes('1.0') |
| 390 | 391 | ? ({ google_search_retrieval: {} }) |
| 391 | 392 | : ({ google_search: {} }); |
| 392 | 393 | tools.push(searchTool); |
| 393 | 394 | } |
| 394 | 395 | |
| 395 | 396 | if (Array.isArray(request.body.tools) && request.body.tools.length > 0 && !useMultiModal && !isGemma) { |
| 396 | 397 | const functionDeclarations = []; |
| 397 | 398 | for (const tool of request.body.tools) { |
| 398 | 399 | if (tool.type === 'function') { |