GradScaler
==========

Utility for gradient scaling in mixed precision training.

**Abstract Signature:**

``GradScaler(init_scale: float = 65536.0, growth_factor: float = 2.0, backoff_factor: float = 0.5, growth_interval: int = 2000, enabled: bool = True)``

.. raw:: html

 
    <div class="op-tabs-container"> 
      <div class="op-tabs-nav"> 
        <button class="op-tab-btn active" onclick="openOpTab(event, 'PyTorch_0')">PyTorch</button><button class="op-tab-btn " onclick="openOpTab(event, 'JAX (Core)_1')">JAX (Core)</button><button class="op-tab-btn " onclick="openOpTab(event, 'NumPy_2')">NumPy</button><button class="op-tab-btn " onclick="openOpTab(event, 'Keras_3')">Keras</button><button class="op-tab-btn " onclick="openOpTab(event, 'TensorFlow_4')">TensorFlow</button><button class="op-tab-btn " onclick="openOpTab(event, 'Apple MLX_5')">Apple MLX</button><button class="op-tab-btn " onclick="openOpTab(event, 'Flax NNX_6')">Flax NNX</button><button class="op-tab-btn " onclick="openOpTab(event, 'PaxML / Praxis_7')">PaxML / Praxis</button> 
      </div> 
      <div class="op-tabs-content"> 
        <div id="PyTorch_0" class="op-tab-pane active"><h4>PyTorch</h4><div class="op-detail-row"><span class="label">API:</span> <code>torch.cuda.amp.GradScaler</code></div><div class="op-detail-row"><span class="label">Strategy:</span> <span>Direct Mapping</span></div><div class="op-detail-row"><a href="https://pytorch.org/docs/stable/generated/torch.cuda.amp.GradScaler.html" target="_blank" class="op-doc-link">Official Docs ↗</a></div></div><div id="JAX (Core)_1" class="op-tab-pane "><h4>JAX (Core)</h4><div class="op-detail-row"><span class="label">API:</span> <code>optax.amp.GradScaler</code></div><div class="op-detail-row"><span class="label">Strategy:</span> <span>Direct Mapping</span></div><div class="op-detail-row"><a href="https://jax.readthedocs.io/en/latest/_autosummary/optax.amp.GradScaler.html" target="_blank" class="op-doc-link">Official Docs ↗</a></div></div><div id="NumPy_2" class="op-tab-pane "><h4>NumPy</h4><div class="op-detail-row"><span class="label">API:</span> <code>—</code></div><div class="op-detail-row"><span class="label">Strategy:</span> <span>Custom / Partial</span></div></div><div id="Keras_3" class="op-tab-pane "><h4>Keras</h4><div class="op-detail-row"><span class="label">API:</span> <code>keras.mixed_precision.LossScaleOptimizer</code></div><div class="op-detail-row"><span class="label">Strategy:</span> <span>Direct Mapping</span></div><div class="op-detail-row"><a href="https://keras.io/search.html?q=keras.mixed_precision.LossScaleOptimizer" target="_blank" class="op-doc-link">Official Docs ↗</a></div></div><div id="TensorFlow_4" class="op-tab-pane "><h4>TensorFlow</h4><div class="op-detail-row"><span class="label">API:</span> <code>tf.keras.mixed_precision.LossScaleOptimizer</code></div><div class="op-detail-row"><span class="label">Strategy:</span> <span>Direct Mapping</span></div><div class="op-detail-row"><a href="https://www.tensorflow.org/api_docs/python/tf/keras/mixed_precision/LossScaleOptimizer" target="_blank" class="op-doc-link">Official Docs ↗</a></div></div><div id="Apple MLX_5" class="op-tab-pane "><h4>Apple MLX</h4><div class="op-detail-row"><span class="label">API:</span> <code>—</code></div><div class="op-detail-row"><span class="label">Strategy:</span> <span>Custom / Partial</span></div></div><div id="Flax NNX_6" class="op-tab-pane "><h4>Flax NNX</h4><div class="op-detail-row"><span class="label">API:</span> <code>optax.amp.GradScaler</code></div><div class="op-detail-row"><span class="label">Strategy:</span> <span>Direct Mapping</span></div><div class="op-detail-row"><a href="https://flax.readthedocs.io/en/latest/search.html?q=optax.amp.GradScaler" target="_blank" class="op-doc-link">Official Docs ↗</a></div></div><div id="PaxML / Praxis_7" class="op-tab-pane "><h4>PaxML / Praxis</h4><div class="op-detail-row"><span class="label">API:</span> <code>praxis.optimizers.ShardedAdafactor</code></div><div class="op-detail-row"><span class="label">Strategy:</span> <span>Direct Mapping</span></div><div class="op-detail-row"><a href="https://github.com/search?q=repo%3Agoogle%2Fpaxml+praxis.optimizers.ShardedAdafactor&type=code" target="_blank" class="op-doc-link">Official Docs ↗</a></div></div> 
      </div> 

      <!-- Load JS Logic only once per page ideally, but safe to exist globally --> 
      <script src="../_static/op_tabs.js"></script> 
    </div>