Merge pull request #3537 from Kas1o/staging Updated Groq model options to sync with official playground

1a1afd00a1f2b2ad1795397d1c779eec63fa3932

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

Signed
Parents: 7188060ac8d2e6c5f4e529f480d444d301d37911 e7d38d95d0df4b1bbf1a19d1a61533bb76288b00
3 files changed, +64 -40Ignore whitespace
public/index.html+25 -13
@@ -3186,21 +3186,33 @@
3186 </div>3186 </div>
3187 <h4 data-i18n="Groq Model">Groq Model</h4>3187 <h4 data-i18n="Groq Model">Groq Model</h4>
3188 <select id="model_groq_select">3188 <select id="model_groq_select">
3189 <optgroup label="Production Models">3189 <optgroup label="Alibaba Cloud">
3190 <option value="gemma2-9b-it">gemma2-9b-it</option>3190 <option value="qwen-2.5-32b">qwen-2.5-32b</option>
3191 <option value="llama-3.3-70b-versatile">llama-3.3-70b-versatile</option>3191 <option value="qwen-2.5-coder-32b">qwen-2.5-coder-32b</option>
3192 <option value="llama-3.1-8b-instant">llama-3.1-8b-instant</option>
3193 <option value="llama3-70b-8192">llama3-70b-8192</option>
3194 <option value="llama3-8b-8192">llama3-8b-8192</option>
3195 <option value="mixtral-8x7b-32768">mixtral-8x7b-32768</option>
3196 </optgroup>3192 </optgroup>
3197 <optgroup label="Preview Models">3193 <optgroup label="DeepSeek / Alibaba Cloud">
3194 <option value="deepseek-r1-distill-qwen-32b">deepseek-r1-distill-qwen-32b</option>
3195 </optgroup>
3196 <optgroup label="DeepSeek / Meta">
3198 <option value="deepseek-r1-distill-llama-70b">deepseek-r1-distill-llama-70b</option>3197 <option value="deepseek-r1-distill-llama-70b">deepseek-r1-distill-llama-70b</option>
3199 <option value="llama-3.3-70b-specdec">llama-3.3-70b-specdec</option>3198 </optgroup>
3200 <option value="llama-3.2-1b-preview">llama-3.2-1b-preview</option>3199 <optgroup label="Google">
3201 <option value="llama-3.2-3b-preview">llama-3.2-3b-preview</option>3200 <option value="gemma2-9b-it">gemma2-9b-it</option>
3202 <option value="llama-3.2-11b-vision-preview">llama-3.2-11b-vision-preview</option>3201 </optgroup>
3203 <option value="llama-3.2-90b-vision-preview">llama-3.2-90b-vision-preview</option>3202 <optgroup label="Meta">
3203 <option value="llama-3.1-8b-instant">llama-3.1-8b-instant </option>
3204 <option value="llama-3.2-11b-vision-preview">llama-3.2-11b-vision-preview </option>
3205 <option value="llama-3.2-1b-preview">llama-3.2-1b-preview </option>
3206 <option value="llama-3.2-3b-preview">llama-3.2-3b-preview </option>
3207 <option value="llama-3.2-90b-vision-preview">llama-3.2-90b-vision-preview </option>
3208 <option value="llama-3.3-70b-specdec">llama-3.3-70b-specdec </option>
3209 <option value="llama-3.3-70b-versatile">llama-3.3-70b-versatile </option>
3210 <option value="llama-guard-3-8b">llama-guard-3-8b </option>
3211 <option value="llama3-70b-8192">llama3-70b-8192 </option>
3212 <option value="llama3-8b-8192">llama3-8b-8192 </option>
3213 </optgroup>
3214 <optgroup label="Mistral AI">
3215 <option value="mixtral-8x7b-32768">mixtral-8x7b-32768</option>
3204 </optgroup>3216 </optgroup>
3205 </select>3217 </select>
3206 </div>3218 </div>
public/scripts/openai.js+36 -27
@@ -4115,6 +4115,40 @@ function getMaxContextWindowAI(value) {
4115 }4115 }
4116}4116}
41174117
4118/**
4119 * Get the maximum context size for the Groq model
4120 * @param {string} model Model identifier
4121 * @param {boolean} isUnlocked Whether context limits are unlocked
4122 * @returns {number} Maximum context size in tokens
4123 */
4124function getGroqMaxContext(model, isUnlocked) {
4125 if (isUnlocked) {
4126 return unlocked_max;
4127 }
4128
4129 const contextMap = {
4130 'gemma2-9b-it': max_8k,
4131 'llama-3.3-70b-versatile': max_128k,
4132 'llama-3.1-8b-instant': max_128k,
4133 'llama3-70b-8192': max_8k,
4134 'llama3-8b-8192': max_8k,
4135 'llama-guard-3-8b': max_8k,
4136 'mixtral-8x7b-32768': max_32k,
4137 'deepseek-r1-distill-llama-70b': max_128k,
4138 'llama-3.3-70b-specdec': max_8k,
4139 'llama-3.2-1b-preview': max_128k,
4140 'llama-3.2-3b-preview': max_128k,
4141 'llama-3.2-11b-vision-preview': max_128k,
4142 'llama-3.2-90b-vision-preview': max_128k,
4143 'qwen-2.5-32b': max_128k,
4144 'deepseek-r1-distill-qwen-32b': max_128k,
4145 'deepseek-r1-distill-llama-70b-specdec': max_128k,
4146 };
4147
4148 // Return context size if model found, otherwise default to 128k
4149 return Object.entries(contextMap).find(([key]) => model.includes(key))?.[1] || max_128k;
4150}
4151
4118async function onModelChange() {4152async function onModelChange() {
4119 biasCache = undefined;4153 biasCache = undefined;
4120 let value = String($(this).val() || '');4154 let value = String($(this).val() || '');
@@ -4416,33 +4450,8 @@ async function onModelChange() {
4416 }4450 }
44174451
4418 if (oai_settings.chat_completion_source == chat_completion_sources.GROQ) {4452 if (oai_settings.chat_completion_source == chat_completion_sources.GROQ) {
4419 if (oai_settings.max_context_unlocked) {4453 const maxContext = getGroqMaxContext(oai_settings.groq_model, oai_settings.max_context_unlocked);
4420 $('#openai_max_context').attr('max', unlocked_max);4454 $('#openai_max_context').attr('max', maxContext);
4421 } else if (oai_settings.groq_model.includes('gemma2-9b-it')) {
4422 $('#openai_max_context').attr('max', max_8k);
4423 } else if (oai_settings.groq_model.includes('llama-3.3-70b-versatile')) {
4424 $('#openai_max_context').attr('max', max_128k);
4425 } else if (oai_settings.groq_model.includes('llama-3.1-8b-instant')) {
4426 $('#openai_max_context').attr('max', max_128k);
4427 } else if (oai_settings.groq_model.includes('llama3-70b-8192')) {
4428 $('#openai_max_context').attr('max', max_8k);
4429 } else if (oai_settings.groq_model.includes('llama3-8b-8192')) {
4430 $('#openai_max_context').attr('max', max_8k);
4431 } else if (oai_settings.groq_model.includes('mixtral-8x7b-32768')) {
4432 $('#openai_max_context').attr('max', max_32k);
4433 } else if (oai_settings.groq_model.includes('deepseek-r1-distill-llama-70b')) {
4434 $('#openai_max_context').attr('max', max_128k);
4435 } else if (oai_settings.groq_model.includes('llama-3.3-70b-specdec')) {
4436 $('#openai_max_context').attr('max', max_8k);
4437 } else if (oai_settings.groq_model.includes('llama-3.2-1b-preview')) {
4438 $('#openai_max_context').attr('max', max_128k);
4439 } else if (oai_settings.groq_model.includes('llama-3.2-3b-preview')) {
4440 $('#openai_max_context').attr('max', max_128k);
4441 } else if (oai_settings.groq_model.includes('llama-3.2-11b-vision-preview')) {
4442 $('#openai_max_context').attr('max', max_128k);
4443 } else if (oai_settings.groq_model.includes('llama-3.2-90b-vision-preview')) {
4444 $('#openai_max_context').attr('max', max_128k);
4445 }
4446 oai_settings.openai_max_context = Math.min(Number($('#openai_max_context').attr('max')), oai_settings.openai_max_context);4455 oai_settings.openai_max_context = Math.min(Number($('#openai_max_context').attr('max')), oai_settings.openai_max_context);
4447 $('#openai_max_context').val(oai_settings.openai_max_context).trigger('input');4456 $('#openai_max_context').val(oai_settings.openai_max_context).trigger('input');
4448 oai_settings.temp_openai = Math.min(oai_max_temp, oai_settings.temp_openai);4457 oai_settings.temp_openai = Math.min(oai_max_temp, oai_settings.temp_openai);
public/scripts/tokenizers.js+3 -0
@@ -694,6 +694,9 @@ export function getTokenizerModel() {
694 }694 }
695695
696 if (oai_settings.chat_completion_source === chat_completion_sources.GROQ) {696 if (oai_settings.chat_completion_source === chat_completion_sources.GROQ) {
697 if (oai_settings.groq_model.includes('qwen')) {
698 return qwen2Tokenizer;
699 }
697 if (oai_settings.groq_model.includes('llama-3') || oai_settings.groq_model.includes('llama3')) {700 if (oai_settings.groq_model.includes('llama-3') || oai_settings.groq_model.includes('llama3')) {
698 return llama3Tokenizer;701 return llama3Tokenizer;
699 }702 }