Update QeRL content to clarify hardware support for NVFP4 by removing reference to NF4 format, ensuring accurate representation of the quantization method used in experiments, and emphasizing the efficiency gains achieved through optimized kernels.

Author: vukrosic

app/blog/qerl-quantization-reinforcement-learning/page.tsx

@@ -31,11 +31,11 @@ export default function QeRLProject() {
           const frontmatterContent = frontmatterMatch[1];
           const markdownBody = frontmatterMatch[2];
 
-          // Parse YAML-like frontmatter (simple parsing for our use case)
+          // Parse YAML-like frontmatter
           const heroData: HeroData = {
-            title: "QeRL: Beyond Efficiency",
-            subtitle: "Quantization-enhanced Reinforcement Learning for LLMs",
-            tags: ["⏱️ Technical Deep Dive", "📄 Research Article"]
+            title: "",
+            subtitle: "",
+            tags: []
           };
 
           // Extract values from frontmatter
@@ -47,16 +47,24 @@ export default function QeRLProject() {
             const trimmedLine = line.trim();
             if (trimmedLine.startsWith('hero:')) continue;
 
-            if (trimmedLine.includes(':')) {
-              const [key, ...valueParts] = trimmedLine.split(':');
-              const value = valueParts.join(':').trim().replace(/^["']|["']$/g, '');
+            if (trimmedLine.includes(':') && !trimmedLine.startsWith('-')) {
+              // Save previous array if exists
+              if (currentKey === 'tags' && currentArray.length > 0) {
+                heroData.tags = currentArray;
+              }
+              
+              const colonIndex = trimmedLine.indexOf(':');
+              const key = trimmedLine.substring(0, colonIndex).trim();
+              const value = trimmedLine.substring(colonIndex + 1).trim().replace(/^["']|["']$/g, '');
 
-              switch (key.trim()) {
+              switch (key) {
                 case 'title':
                   heroData.title = value;
+                  currentKey = '';
                   break;
                 case 'subtitle':
                   heroData.subtitle = value;
+                  currentKey = '';
                   break;
                 case 'tags':
                   currentKey = 'tags';
@@ -65,20 +73,14 @@ export default function QeRLProject() {
               }
             } else if (trimmedLine.startsWith('- ')) {
               if (currentKey === 'tags') {
-                const tagValue = trimmedLine.substring(2).replace(/^["']|["']$/g, '');
+                const tagValue = trimmedLine.substring(2).trim().replace(/^["']|["']$/g, '');
                 currentArray.push(tagValue);
               }
-            } else if (trimmedLine === '' && currentArray.length > 0) {
-              if (currentKey === 'tags') {
-                heroData.tags = currentArray;
-                currentArray = [];
-                currentKey = '';
-              }
             }
           }
 
           // Handle final array
-          if (currentArray.length > 0 && currentKey === 'tags') {
+          if (currentKey === 'tags' && currentArray.length > 0) {
             heroData.tags = currentArray;
           }
 
@@ -154,11 +156,11 @@ export default function QeRLProject() {
             <div className="relative">
               <h1 className="text-4xl md:text-5xl lg:text-6xl font-medium mb-8 leading-tight">
                 <span className="bg-gradient-to-r from-blue-400 via-purple-400 to-cyan-400 bg-clip-text text-transparent">
-                  {heroData?.title || 'QeRL: Beyond Efficiency'}
+                  {heroData?.title || 'Train 32B LLM Reasoning On 1 GPU - H100 80GB - QeRL'}
                 </span>
               </h1>
               <div className="text-lg md:text-xl text-slate-400 mb-8">
-                {heroData?.subtitle || 'Quantization-enhanced Reinforcement Learning for LLMs'}
+                {heroData?.subtitle || 'LLM Reinforcement Learning With 4Bit Quantization'}
               </div>
 
               {/* Tags */}
@@ -220,7 +222,7 @@ export default function QeRLProject() {
               {/* Glow effect for the title */}
               <div className="absolute inset-0 text-4xl md:text-5xl lg:text-6xl font-medium leading-tight blur-sm pointer-events-none">
                 <span className="bg-gradient-to-r from-blue-400/20 via-purple-400/20 to-cyan-400/20 bg-clip-text text-transparent">
-                  {heroData?.title || 'QeRL: Beyond Efficiency'}
+                  {heroData?.title || 'Train 32B LLM Reasoning On 1 GPU - H100 80GB - QeRL'}
                 </span>
               </div>
             </div>