adding Automatic mixed precision quantization documentation under Advanced Features

Author: Abhijeet

docs/_build/html/.doctrees/environment.pickle

@@ -43,7 +43,7 @@ Fully automatic, Hessian-aware per-layer bit width assignment using a
 greedy algorithm. Enabled by setting ``auto_quantization: True``:
 
 * Estimates per-layer sensitivity via Hessian eigenvalue (power iteration)
-* Greedy assignment from ``{2, 4, 8}`` bit widths maximising accuracy per bit
+* Greedy assignment from ``{2, 4, 8, 32}`` bit widths maximising accuracy per bit
 * Automatic average bit width selection via binary search calibration
 * Fixes regression tasks where uniform 8-bit QAT fails
 

docs/_build/html/.doctrees/features/index.doctree

@@ -1420,7 +1420,7 @@ <h2><span class="section-number">8.8. </span>Feature Overview<a class="headerlin
 greedy algorithm. Enabled by setting <code class="docutils literal notranslate"><span class="pre">auto_quantization:</span> <span class="pre">True</span></code>:</p>
 <ul class="simple">
 <li><p>Estimates per-layer sensitivity via Hessian eigenvalue (power iteration)</p></li>
-<li><p>Greedy assignment from <code class="docutils literal notranslate"><span class="pre">{2,</span> <span class="pre">4,</span> <span class="pre">8}</span></code> bit widths maximising accuracy per bit</p></li>
+<li><p>Greedy assignment from <code class="docutils literal notranslate"><span class="pre">{2,</span> <span class="pre">4,</span> <span class="pre">8,</span> <span class="pre">32}</span></code> bit widths maximising accuracy per bit</p></li>
 <li><p>Automatic average bit width selection via binary search calibration</p></li>
 <li><p>Fixes regression tasks where uniform 8-bit QAT fails</p></li>
 </ul>

docs/_build/html/_sources/features/index.rst.txt

@@ -43,7 +43,7 @@ Fully automatic, Hessian-aware per-layer bit width assignment using a
 greedy algorithm. Enabled by setting ``auto_quantization: True``:
 
 * Estimates per-layer sensitivity via Hessian eigenvalue (power iteration)
-* Greedy assignment from ``{2, 4, 8}`` bit widths maximising accuracy per bit
+* Greedy assignment from ``{2, 4, 8, 32}`` bit widths maximising accuracy per bit
 * Automatic average bit width selection via binary search calibration
 * Fixes regression tasks where uniform 8-bit QAT fails