Update README.md

README.md

@@ -1,3 +1,267 @@
 ---
 license: apache-2.0
 ---
+
+<!-- 
+  NORTH CODE QUANT - MODEL CARD TEMPLATE
+  Designed for GGUF Quantizations
+-->
+
+<style>
+  :root {
+    --nc-bg: #0f1115;
+    --nc-card: #161b22;
+    --nc-border: #30363d;
+    --nc-text-main: #e6edf3;
+    --nc-text-muted: #8b949e;
+    --nc-accent: #58a6ff;
+    --nc-accent-glow: rgba(88, 166, 255, 0.15);
+    --nc-code-bg: #0d1117;
+    --font-sans: -apple-system, BlinkMacSystemFont, "Segoe UI", Helvetica, Arial, sans-serif;
+    --font-mono: ui-monospace, SFMono-Regular, "SF Mono", Menlo, Consolas, monospace;
+  }
+
+  .nc-container {
+    font-family: var(--font-sans);
+    color: var(--nc-text-main);
+    line-height: 1.6;
+    max-width: 900px;
+    margin: 0 auto;
+    padding: 2rem 1rem;
+    background-color: var(--nc-bg);
+  }
+
+  /* Typography & Links */
+  .nc-container h1 { font-size: 2.5rem; border-bottom: none; margin-bottom: 0.5rem; letter-spacing: -0.02em; }
+  .nc-container h2 { font-size: 1.5rem; margin-top: 2.5rem; border-bottom: 1px solid var(--nc-border); padding-bottom: 0.5rem; }
+  .nc-container h3 { font-size: 1.1rem; color: var(--nc-text-muted); text-transform: uppercase; letter-spacing: 0.05em; margin-top: 2rem;}
+  .nc-container a { color: var(--nc-accent); text-decoration: none; transition: opacity 0.2s; }
+  .nc-container a:hover { text-decoration: underline; }
+  
+  /* Hero Section */
+  .nc-hero {
+    display: flex;
+    align-items: center;
+    gap: 1rem;
+    margin-bottom: 2rem;
+  }
+  .nc-badge {
+    background: var(--nc-accent-glow);
+    color: var(--nc-accent);
+    padding: 0.25rem 0.75rem;
+    border-radius: 999px;
+    font-size: 0.85rem;
+    font-weight: 600;
+    border: 1px solid rgba(88, 166, 255, 0.3);
+  }
+
+  /* Info Grid */
+  .nc-grid {
+    display: grid;
+    grid-template-columns: repeat(auto-fit, minmax(200px, 1fr));
+    gap: 1rem;
+    margin: 1.5rem 0;
+  }
+  .nc-stat-card {
+    background: var(--nc-card);
+    border: 1px solid var(--nc-border);
+    border-radius: 8px;
+    padding: 1rem;
+  }
+  .nc-stat-label { font-size: 0.8rem; color: var(--nc-text-muted); display: block; margin-bottom: 0.25rem; }
+  .nc-stat-value { font-size: 1.1rem; font-weight: 600; color: var(--nc-text-main); }
+
+  /* Code Blocks */
+  .nc-code-block {
+    position: relative;
+    background: var(--nc-code-bg);
+    border: 1px solid var(--nc-border);
+    border-radius: 8px;
+    padding: 1rem;
+    margin: 1rem 0;
+    overflow-x: auto;
+  }
+  .nc-code-block code {
+    font-family: var(--font-mono);
+    font-size: 0.9rem;
+    color: #c9d1d9;
+    white-space: pre;
+  }
+  .nc-cli-label {
+    position: absolute;
+    top: 0.5rem;
+    right: 0.75rem;
+    font-size: 0.7rem;
+    color: var(--nc-text-muted);
+    text-transform: uppercase;
+    letter-spacing: 0.05em;
+    pointer-events: none;
+  }
+
+  /* GGUF Table */
+  .nc-table-wrapper { overflow-x: auto; margin: 1.5rem 0; border-radius: 8px; border: 1px solid var(--nc-border); }
+  table.nc-table { width: 100%; border-collapse: collapse; font-size: 0.9rem; }
+  table.nc-table th { 
+    background: var(--nc-card); 
+    text-align: left; 
+    padding: 0.75rem 1rem; 
+    color: var(--nc-text-muted); 
+    font-weight: 600; 
+    border-bottom: 1px solid var(--nc-border);
+  }
+  table.nc-table td { padding: 0.75rem 1rem; border-bottom: 1px solid var(--nc-border); }
+  table.nc-table tr:last-child td { border-bottom: none; }
+  table.nc-table tr:hover td { background: rgba(255,255,255,0.02); }
+  .nc-quant-tag { 
+    font-family: var(--font-mono); 
+    font-size: 0.8rem; 
+    background: rgba(255,255,255,0.06); 
+    padding: 0.15rem 0.4rem; 
+    border-radius: 4px; 
+  }
+
+  /* Disclaimer Box */
+  .nc-disclaimer {
+    margin-top: 3rem;
+    padding: 1rem;
+    border-left: 3px solid var(--nc-border);
+    color: var(--nc-text-muted);
+    font-size: 0.85rem;
+  }
+
+  /* Responsive */
+  @media (max-width: 600px) {
+    .nc-container h1 { font-size: 1.8rem; }
+    .nc-grid { grid-template-columns: 1fr 1fr; }
+  }
+</style>
+
+<div class="nc-container">
+
+  <!-- HERO -->
+  <div class="nc-hero">
+    <h1>North Code Quant</h1>
+    <span class="nc-badge">GGUF</span>
+    <span class="nc-badge">Code Generation</span>
+  </div>
+  <p style="color: var(--nc-text-muted); font-size: 1.1rem; margin-top: -1rem;">
+    High-performance quantized GGUF builds of Cohere's North Code model.<br>
+    Optimized for local inference via llama.cpp, LM Studio, and Ollama.
+  </p>
+
+  <!-- KEY STATS -->
+  <div class="nc-grid">
+    <div class="nc-stat-card">
+      <span class="nc-stat-label">Base Model</span>
+      <span class="nc-stat-value">Cohere North Code</span>
+    </div>
+    <div class="nc-stat-card">
+      <span class="nc-stat-label">Architecture</span>
+      <span class="nc-stat-value">Cohere / Command-R</span>
+    </div>
+    <div class="nc-stat-card">
+      <span class="nc-stat-label">Context Length</span>
+      <span class="nc-stat-value">128K Tokens</span>
+    </div>
+    <div class="nc-stat-card">
+      <span class="nc-stat-label">License</span>
+      <span class="nc-stat-value">CC-BY-NC / Custom</span>
+    </div>
+  </div>
+
+  <!-- QUICK START -->
+  <h2>⚡ Quick Start</h2>
+  
+  <h3>Ollama</h3>
+  <div class="nc-code-block">
+    <span class="nc-cli-label">CLI</span>
+    <code>ollama run north-code-quant</code>
+  </div>
+
+  <h3>LM Studio</h3>
+  <p style="color: var(--nc-text-muted); font-size: 0.95rem;">
+    Search for <strong>"North Code Quant"</strong> in the LM Studio search bar, select your preferred quantization level from the sidebar, and click Download.
+  </p>
+
+  <h3>llama.cpp</h3>
+  <div class="nc-code-block">
+    <span class="nc-cli-label">Bash</span>
+    <code>./llama-cli -m north-code-quant-Q4_K_M.gguf \
+  --ctx-size 8192 \
+  --threads $(nproc) \
+  --prompt "def fibonacci(n):"</code>
+  </div>
+
+  <!-- AVAILABLE QUANTS -->
+  <h2>📦 Available Quants</h2>
+  <p style="color: var(--nc-text-muted); margin-bottom: 1rem;">
+    Files are sorted by size and quality. <strong>Q4_K_M</strong> is recommended for most users as the best balance of speed and perplexity.
+  </p>
+
+  <div class="nc-table-wrapper">
+    <table class="nc-table">
+      <thead>
+        <tr>
+          <th>File Name</th>
+          <th>Quant Type</th>
+          <th>Size</th>
+          <th>Description</th>
+        </tr>
+      </thead>
+      <tbody>
+        <!-- UPDATE THESE ROWS WITH YOUR ACTUAL FILES -->
+        <tr>
+          <td><code>north-code-quant-Q8_0.gguf</code></td>
+          <td><span class="nc-quant-tag">Q8_0</span></td>
+          <td>-- GB</td>
+          <td>Near-lossless. Best quality, higher VRAM/RAM requirement.</td>
+        </tr>
+        <tr>
+          <td><code>north-code-quant-Q6_K.gguf</code></td>
+          <td><span class="nc-quant-tag">Q6_K</span></td>
+          <td>-- GB</td>
+          <td>Very low perplexity loss. Great for critical code tasks.</td>
+        </tr>
+        <tr style="background: rgba(88, 166, 255, 0.05);">
+          <td><code>north-code-quant-Q4_K_M.gguf</code></td>
+          <td><span class="nc-quant-tag">Q4_K_M</span></td>
+          <td>-- GB</td>
+          <td><strong>⭐ Recommended.</strong> Best size/performance ratio.</td>
+        </tr>
+        <tr>
+          <td><code>north-code-quant-Q3_K_M.gguf</code></td>
+          <td><span class="nc-quant-tag">Q3_K_M</span></td>
+          <td>-- GB</td>
+          <td>Lower resource usage. Noticeable quality degradation in complex reasoning.</td>
+        </tr>
+        <tr>
+          <td><code>north-code-quant-IQ4_XS.gguf</code></td>
+          <td><span class="nc-quant-tag">IQ4_XS</span></td>
+          <td>-- GB</td>
+          <td>Importance Matrix quant. Smaller than Q4 with similar quality.</td>
+        </tr>
+      </tbody>
+    </table>
+  </div>
+
+  <!-- MODEL DETAILS -->
+  <h2>📝 About This Quantization</h2>
+  <p>
+    These GGUF files were converted from the official 
+    <a href="https://huggingface.co/CohereForAI/c4ai-command-r-plus" target="_blank">Cohere North Code</a> 
+    weights using <code>llama.cpp</code> with importance matrix calibration for optimal token-level precision retention.
+  </p>
+  <ul style="color: var(--nc-text-muted); padding-left: 1.5rem;">
+    <li><strong>Calibration Dataset:</strong> CodeAlpaca + SlimOrca subset</li>
+    <li><strong>Imatrix:</strong> Enabled for K-quants (Q3_K through Q6_K)</li>
+    <li><strong>Vocab:</strong> Original Cohere tokenizer preserved</li>
+  </ul>
+
+  <!-- DISCLAIMER -->
+  <div class="nc-disclaimer">
+    <strong>⚠️ Disclaimer:</strong> This is a quantized derivative model. While quants retain most of the base model's capabilities, 
+    lower-bit quantizations may exhibit degraded performance in edge-case code generation or multilingual tasks. 
+    Always verify generated code before execution. This model inherits the license terms of the original Cohere North Code model.
+  </div>
+
+</div>