add Qwen 1.5 0.5B and 1.8B

index.html

@@ -350,8 +350,10 @@ <h6>Model Selection</h6>
                                                             Xenova/jina-embeddings-v2-base-de | 💾162Mb 📥0 ❤️0</option>
                                                         <option value="Xenova/jina-embeddings-v2-base-zh">
                                                             Xenova/jina-embeddings-v2-base-zh | 💾162Mb 📥0 ❤️0</option>
-                                                        <option value="jinaai/jina-embeddings-v2-base-code">
+                                                         <!--<option value="jinaai/jina-embeddings-v2-base-code">
                                                             jinaai/jina-embeddings-v2-base-code | 💾162Mb 📥0 ❤️0</option>
+                                                        <option value="jinaai/jina-embeddings-v2-base-es">
+                                                            jinaai/jina-embeddings-v2-base-es | 💾162Mb 📥0 ❤️0</option> --> <!-- access control inhibits direct loading from HF -->
                                                         <option value="do-me/jina-embeddings-v2-small-en">
                                                             do-me/jina-embeddings-v2-small-en (test)</option>
                                                         <option value="do-me/jina-embeddings-v2-base-en">
@@ -728,8 +730,10 @@ <h4>Chat (Retrieval Augmented Generation, RAG)</h4>
                                 <div class="col-md-8">
                                     <div class="form-floating mb-2">
                                         <select class="form-select form-control" id="chat-model-name">
-                                            <option value="Xenova/Qwen1.5-0.5B-Chat">Xenova/Qwen1.5-0.5B-Chat</option>
-                                            <option selected value="Xenova/LaMini-Flan-T5-783M">
+                                            <option selected value="Xenova/Qwen1.5-0.5B-Chat">Xenova/Qwen1.5-0.5B-Chat (0.482Gb)</option>
+                                            <option value="Xenova/Qwen1.5-1.8B-Chat">Xenova/Qwen1.5-1.8B-Chat (1.87Gb)</option>
+
+                                            <option value="Xenova/LaMini-Flan-T5-783M">
                                                 Xenova/LaMini-Flan-T5-783M | 💾1.65Gb 📥17 ❤️20</option>
                                             <option value="Xenova/t5-small">Xenova/t5-small | 💾81Mb 📥2
                                                 ❤️2</option>
@@ -772,7 +776,7 @@ <h4>Chat (Retrieval Augmented Generation, RAG)</h4>
                             </div>
 
                         </div>
-                        <p id="chat_text" class="ml-2"></p>
+                        <div id="chat_text" class="ml-2"></div>
                         <!-- progress bar -->
                         <div class="col-12">
                             <div class="progress" role="progressbar" aria-label="Loading model" aria-valuenow="0"

package.json

@@ -15,6 +15,7 @@
     "bootstrap": "^5.3.2",
     "codemirror": "^5.52.2",
     "deck.gl": "^8.9.34",
+    "marked": "^12.0.0",
     "pako": "^2.1.0",
     "wasm-bhtsne": "^0.3.3"
   },

src/js/semantic.js

@@ -38,7 +38,6 @@ let loadResolve;
  */
 let queryResolve;
 
-
 function downloadFile(data, filename, mimeType) {
     const blob = new Blob([data], { type: mimeType });
 

src/js/worker.js

@@ -1,6 +1,8 @@
 import { pipeline, AutoTokenizer } from '@xenova/transformers';
 import pako from 'pako';
 import init, { tSNE } from "wasm-bhtsne";
+import { marked } from 'marked';
+
 init();
 // env.useBrowserCache = false; // for testing
 
@@ -81,36 +83,38 @@ async function getTokens(text) {
 async function chat(text, max_new_tokens = 100) {
     return new Promise(async (resolve, reject) => {
         // hier Weiche einbauen für Qwen da tokenizer anders
-        console.log(chat_model_name);
+        console.log(chat_model_name, max_new_tokens);
 
         if (chat_model_name.includes("Qwen")) {
             try {
 
                 // Define the prompt and list of messages
-
                 const messages = [
                     { "role": "system", "content": "You are a helpful assistant." },
                     { "role": "user", "content": text }
                 ]
 
                 const generatorText = chat_generator.tokenizer.apply_chat_template(messages, {
                     tokenize: false,
-                    add_generation_prompt: true,
+                    add_generation_prompt: false,
                 });
 
                 const thisChat = await chat_generator(generatorText, {
                     max_new_tokens: max_new_tokens,
                     do_sample: false,
                     callback_function: async function (beams) {
                         //const decodedText = await token_to_text(beams, chat_generator.tokenizer);
-                        //console.log(decodedText);
-                        console.log(beams)
+                        let decodedText = chat_generator.tokenizer.decode(beams[0].output_token_ids, { skip_special_tokens: false })
+
+                        decodedText = decodedText.split("<|im_start|>")[3].replace("<|im_end|>","") // just return the model's output
+                        decodedText = marked(decodedText)
+
                         self.postMessage({
                             type: 'chat',
-                            chat_text: beams,
+                            chat_text: decodedText
                         });
 
-                        resolve(beams); // Resolve the main promise with chat text
+                        resolve(decodedText); // Resolve the main promise with chat text
                     },
                 });
             } catch (error) {