Spaces:

webml-community
/

TranslateGemma-WebGPU

Running

nico-martin HF Staff commited on 4 days ago

Commit

82ca2ac

1 Parent(s): 5ba40ca

changed waiting text

Files changed (3) hide show

dist/assets/{index-C8mlkem9.js → index-Dq-reLEJ.js} RENAMED Viewed

The diff for this file is too large to render. See raw diff

dist/index.html CHANGED Viewed

@@ -43,7 +43,7 @@
       href="https://fonts.googleapis.com/css2?family=Google+Sans:wght@400;500;700&display=swap"
       rel="stylesheet"
     />
-    <script type="module" crossorigin src="/assets/index-C8mlkem9.js"></script>
     <link rel="stylesheet" crossorigin href="/assets/index-DcjTpXS7.css">
   </head>
   <body>

       href="https://fonts.googleapis.com/css2?family=Google+Sans:wght@400;500;700&display=swap"
       rel="stylesheet"
     />
+    <script type="module" crossorigin src="/assets/index-Dq-reLEJ.js"></script>
     <link rel="stylesheet" crossorigin href="/assets/index-DcjTpXS7.css">
   </head>
   <body>

src/components/Waiting.tsx CHANGED Viewed

@@ -58,12 +58,12 @@ export default function Waiting({ className = "" }: { className?: string }) {
           {titles[titleIndex]}
         </h3>
         <p className="text-sm text-secondary-foreground">
-          TranslateGemma outperforms classic translation models because it
-          starts from an LLM (Gemma 3) that understands language meaning and
-          world context, not just statistical mappings between word pairs, which
-          gives it better quality on nuanced and low-resource translation
-          without needing massive amounts of parallel training data for every
-          language pair.
         </p>
       </div>

           {titles[titleIndex]}
         </h3>
         <p className="text-sm text-secondary-foreground">
+          TranslateGemma is built on Gemma 3 from Google, and outperforms
+          classic translation models thanks to its strong understanding of
+          language semantics and context, not just statistical mappings between
+          word pairs. This improves quality on nuanced and low-resource
+          translation without needing massive amounts of parallel training data
+          for every language pair.
         </p>
       </div>