diff --git a/docs/SUMMARY.md b/docs/SUMMARY.md
index 7753718d1..1a2fe59f2 100644
--- a/docs/SUMMARY.md
+++ b/docs/SUMMARY.md
@@ -80,6 +80,7 @@
   - [concrete.ml.quantization.md](developer-guide/api/concrete.ml.quantization.md)
   - [concrete.ml.quantization.post_training.md](developer-guide/api/concrete.ml.quantization.post_training.md)
   - [concrete.ml.quantization.quantized_module.md](developer-guide/api/concrete.ml.quantization.quantized_module.md)
+  - [concrete.ml.quantization.quantized_module_passes.md](developer-guide/api/concrete.ml.quantization.quantized_module_passes.md)
   - [concrete.ml.quantization.quantized_ops.md](developer-guide/api/concrete.ml.quantization.quantized_ops.md)
   - [concrete.ml.quantization.quantizers.md](developer-guide/api/concrete.ml.quantization.quantizers.md)
   - [concrete.ml.search_parameters.md](developer-guide/api/concrete.ml.search_parameters.md)
diff --git a/docs/advanced_examples/FullyConnectedNeuralNetwork.ipynb b/docs/advanced_examples/FullyConnectedNeuralNetwork.ipynb
index 574810462..4df7be8ff 100644
--- a/docs/advanced_examples/FullyConnectedNeuralNetwork.ipynb
+++ b/docs/advanced_examples/FullyConnectedNeuralNetwork.ipynb
@@ -11,7 +11,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 1,
+   "execution_count": 14,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -38,7 +38,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": 15,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -62,16 +62,13 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 3,
+   "execution_count": 16,
    "metadata": {},
    "outputs": [],
    "source": [
     "params = {\n",
     "    \"module__n_layers\": 3,\n",
-    "    \"module__n_w_bits\": 3,\n",
-    "    \"module__n_a_bits\": 4,\n",
-    "    \"module__n_accum_bits\": 9,\n",
-    "    \"module__activation_function\": nn.Sigmoid,\n",
+    "    \"module__activation_function\": nn.ReLU,\n",
     "    \"max_epochs\": 1000,\n",
     "    \"verbose\": 0,\n",
     "}\n",
@@ -80,7 +77,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 4,
+   "execution_count": 17,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -89,7 +86,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 5,
+   "execution_count": 18,
    "metadata": {},
    "outputs": [
     {
@@ -110,14 +107,14 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 6,
+   "execution_count": 19,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "The test accuracy of the trained Concrete ML simulated model is 86.84%\n"
+      "The test accuracy of the trained Concrete ML simulated model is 97.37%\n"
      ]
     }
    ],
@@ -138,7 +135,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 7,
+   "execution_count": 20,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -155,14 +152,14 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 8,
+   "execution_count": 21,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Generating a key for a 8-bit circuit\n"
+      "Generating a key for a 9-bit circuit\n"
      ]
     }
    ],
@@ -172,14 +169,14 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 9,
+   "execution_count": 22,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Key generation time: 135.09 seconds\n"
+      "Key generation time: 46.67 seconds\n"
      ]
     }
    ],
@@ -198,21 +195,21 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 10,
+   "execution_count": 23,
    "metadata": {},
    "outputs": [
     {
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "100%|██████████| 38/38 [02:38<00:00,  4.16s/it]"
+      "100%|██████████| 38/38 [00:46<00:00,  1.23s/it]"
      ]
     },
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Execution time: 4.17 seconds per sample\n"
+      "Execution time: 1.23 seconds per sample\n"
      ]
     },
     {
@@ -242,7 +239,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 11,
+   "execution_count": 24,
    "metadata": {},
    "outputs": [
     {
@@ -250,8 +247,8 @@
      "output_type": "stream",
      "text": [
       "Test accuracy using the sklearn model: 100.00%\n",
-      "Test accuracy using the Concrete ML simulated model: 86.84%\n",
-      "Test accuracy using the Concrete ML FHE model: 86.84%\n"
+      "Test accuracy using the Concrete ML simulated model: 97.37%\n",
+      "Test accuracy using the Concrete ML FHE model: 97.37%\n"
      ]
     }
    ],
@@ -280,7 +277,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 12,
+   "execution_count": 25,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -314,12 +311,12 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 13,
+   "execution_count": 26,
    "metadata": {},
    "outputs": [
     {
      "data": {
-      "image/png": "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",
+      "image/png": "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",
       "text/plain": [
        "<Figure size 1600x600 with 2 Axes>"
       ]
diff --git a/docs/developer-guide/api/README.md b/docs/developer-guide/api/README.md
index c2f6cb245..cb8750f4b 100644
--- a/docs/developer-guide/api/README.md
+++ b/docs/developer-guide/api/README.md
@@ -33,6 +33,7 @@
 - [`concrete.ml.quantization.base_quantized_op`](./concrete.ml.quantization.base_quantized_op.md#module-concretemlquantizationbase_quantized_op): Base Quantized Op class that implements quantization for a float numpy op.
 - [`concrete.ml.quantization.post_training`](./concrete.ml.quantization.post_training.md#module-concretemlquantizationpost_training): Post Training Quantization methods.
 - [`concrete.ml.quantization.quantized_module`](./concrete.ml.quantization.quantized_module.md#module-concretemlquantizationquantized_module): QuantizedModule API.
+- [`concrete.ml.quantization.quantized_module_passes`](./concrete.ml.quantization.quantized_module_passes.md#module-concretemlquantizationquantized_module_passes): Optimization passes for QuantizedModules.
 - [`concrete.ml.quantization.quantized_ops`](./concrete.ml.quantization.quantized_ops.md#module-concretemlquantizationquantized_ops): Quantized versions of the ONNX operators for post training quantization.
 - [`concrete.ml.quantization.quantizers`](./concrete.ml.quantization.quantizers.md#module-concretemlquantizationquantizers): Quantization utilities for a numpy array/tensor.
 - [`concrete.ml.search_parameters`](./concrete.ml.search_parameters.md#module-concretemlsearch_parameters): Modules for `p_error` search.
@@ -106,6 +107,7 @@
 - [`post_training.PostTrainingAffineQuantization`](./concrete.ml.quantization.post_training.md#class-posttrainingaffinequantization): Post-training Affine Quantization.
 - [`post_training.PostTrainingQATImporter`](./concrete.ml.quantization.post_training.md#class-posttrainingqatimporter): Converter of Quantization Aware Training networks.
 - [`quantized_module.QuantizedModule`](./concrete.ml.quantization.quantized_module.md#class-quantizedmodule): Inference for a quantized model.
+- [`quantized_module_passes.PowerOfTwoScalingRoundPBSAdapter`](./concrete.ml.quantization.quantized_module_passes.md#class-poweroftwoscalingroundpbsadapter): Detect neural network patterns that can be optimized with round PBS.
 - [`quantized_ops.ONNXConstantOfShape`](./concrete.ml.quantization.quantized_ops.md#class-onnxconstantofshape): ConstantOfShape operator.
 - [`quantized_ops.ONNXGather`](./concrete.ml.quantization.quantized_ops.md#class-onnxgather): Gather operator.
 - [`quantized_ops.ONNXShape`](./concrete.ml.quantization.quantized_ops.md#class-onnxshape): Shape operator.
@@ -280,6 +282,7 @@
 - [`ops_impl.numpy_celu`](./concrete.ml.onnx.ops_impl.md#function-numpy_celu): Compute celu in numpy according to ONNX spec.
 - [`ops_impl.numpy_concatenate`](./concrete.ml.onnx.ops_impl.md#function-numpy_concatenate): Apply concatenate in numpy according to ONNX spec.
 - [`ops_impl.numpy_constant`](./concrete.ml.onnx.ops_impl.md#function-numpy_constant): Return the constant passed as a kwarg.
+- [`ops_impl.numpy_conv`](./concrete.ml.onnx.ops_impl.md#function-numpy_conv): Compute N-D convolution using Torch.
 - [`ops_impl.numpy_cos`](./concrete.ml.onnx.ops_impl.md#function-numpy_cos): Compute cos in numpy according to ONNX spec.
 - [`ops_impl.numpy_cosh`](./concrete.ml.onnx.ops_impl.md#function-numpy_cosh): Compute cosh in numpy according to ONNX spec.
 - [`ops_impl.numpy_div`](./concrete.ml.onnx.ops_impl.md#function-numpy_div): Compute div in numpy according to ONNX spec.
@@ -289,6 +292,7 @@
 - [`ops_impl.numpy_exp`](./concrete.ml.onnx.ops_impl.md#function-numpy_exp): Compute exponential in numpy according to ONNX spec.
 - [`ops_impl.numpy_flatten`](./concrete.ml.onnx.ops_impl.md#function-numpy_flatten): Flatten a tensor into a 2d array.
 - [`ops_impl.numpy_floor`](./concrete.ml.onnx.ops_impl.md#function-numpy_floor): Compute Floor in numpy according to ONNX spec.
+- [`ops_impl.numpy_gemm`](./concrete.ml.onnx.ops_impl.md#function-numpy_gemm): Compute Gemm in numpy according to ONNX spec.
 - [`ops_impl.numpy_greater`](./concrete.ml.onnx.ops_impl.md#function-numpy_greater): Compute greater in numpy according to ONNX spec.
 - [`ops_impl.numpy_greater_float`](./concrete.ml.onnx.ops_impl.md#function-numpy_greater_float): Compute greater in numpy according to ONNX spec and cast outputs to floats.
 - [`ops_impl.numpy_greater_or_equal`](./concrete.ml.onnx.ops_impl.md#function-numpy_greater_or_equal): Compute greater or equal in numpy according to ONNX spec.
diff --git a/docs/developer-guide/api/concrete.ml.onnx.ops_impl.md b/docs/developer-guide/api/concrete.ml.onnx.ops_impl.md
index f0079e5f8..2e8511ed0 100644
--- a/docs/developer-guide/api/concrete.ml.onnx.ops_impl.md
+++ b/docs/developer-guide/api/concrete.ml.onnx.ops_impl.md
@@ -140,7 +140,43 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Constant-13
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L310"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L243"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+
+## <kbd>function</kbd> `numpy_gemm`
+
+```python
+numpy_gemm(
+    a: ndarray,
+    b: ndarray,
+    c: Optional[ndarray] = None,
+    alpha: float = 1,
+    beta: float = 1,
+    transA: int = 0,
+    transB: int = 0
+) → Tuple[ndarray]
+```
+
+Compute Gemm in numpy according to ONNX spec.
+
+See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Gemm-13
+
+**Args:**
+
+- <b>`a`</b> (numpy.ndarray):  Input tensor A. The shape of A should be (M, K) if transA is 0, or (K, M)  if transA is non-zero.
+- <b>`b`</b> (numpy.ndarray):  Input tensor B. The shape of B should be (K, N) if transB is 0, or (N, K)  if transB is non-zero.
+- <b>`c`</b> (Optional\[numpy.ndarray\]):  Optional input tensor C. If not specified, the  computation is done as if C is a scalar 0. The shape of C should be unidirectional  broadcastable to (M, N).  Defaults to None.
+- <b>`alpha`</b> (float):  Scalar multiplier for the product of input tensors A * B.  Defaults to 1.
+- <b>`beta`</b> (float):  Scalar multiplier for input tensor C.  Defaults to 1.
+- <b>`transA`</b> (int):  Whether A should be transposed. The type is kept as int as it is the  type used by ONNX and it can easily be interpreted by Python as a boolean.  Defaults to 0.
+- <b>`transB`</b> (int):  Whether B should be transposed. The type is kept as int as it is the  type used by ONNX and it can easily be interpreted by Python as a boolean.  Defaults to 0.
+
+**Returns:**
+
+- <b>`Tuple[numpy.ndarray]`</b>:  The tuple containing the result tensor
+
+______________________________________________________________________
+
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L309"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_matmul`
 
@@ -163,7 +199,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#MatMul-13
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L328"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L327"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_relu`
 
@@ -185,7 +221,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Relu-14
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L344"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L343"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_sigmoid`
 
@@ -207,7 +243,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sigmoid-13
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L360"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L359"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_softmax`
 
@@ -233,7 +269,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#softmax-13
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L383"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L382"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_cos`
 
@@ -255,7 +291,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Cos-7
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L399"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L398"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_cosh`
 
@@ -277,7 +313,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Cosh-9
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L415"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L414"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_sin`
 
@@ -299,7 +335,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sin-7
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L431"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L430"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_sinh`
 
@@ -321,7 +357,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sinh-9
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L447"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L446"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_tan`
 
@@ -343,7 +379,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Tan-7
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L463"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L462"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_tanh`
 
@@ -365,7 +401,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Tanh-13
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L479"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L478"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_acos`
 
@@ -387,7 +423,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Acos-7
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L495"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L494"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_acosh`
 
@@ -409,7 +445,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Acosh-9
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L511"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L510"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_asin`
 
@@ -431,7 +467,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Asin-7
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L527"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L526"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_asinh`
 
@@ -453,7 +489,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Asinh-9
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L543"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L542"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_atan`
 
@@ -475,7 +511,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Atan-7
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L559"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L558"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_atanh`
 
@@ -497,7 +533,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Atanh-9
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L575"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L574"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_elu`
 
@@ -520,7 +556,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Elu-6
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L591"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L590"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_selu`
 
@@ -548,7 +584,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Selu-6
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L613"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L612"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_celu`
 
@@ -571,7 +607,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Celu-12
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L629"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L628"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_leakyrelu`
 
@@ -594,7 +630,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#LeakyRelu-6
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L645"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L644"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_thresholdedrelu`
 
@@ -617,7 +653,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#ThresholdedRelu-10
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L664"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L663"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_hardsigmoid`
 
@@ -645,7 +681,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#HardSigmoid-6
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L683"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L682"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_softplus`
 
@@ -667,7 +703,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Softplus-1
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L700"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L699"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_abs`
 
@@ -689,7 +725,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Abs-13
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L717"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L716"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_div`
 
@@ -712,7 +748,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Div-14
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L749"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L748"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_mul`
 
@@ -735,7 +771,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Mul-14
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L768"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L767"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_sub`
 
@@ -758,7 +794,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sub-14
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L787"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L786"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_log`
 
@@ -780,7 +816,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Log-13
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L830"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L829"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_erf`
 
@@ -802,7 +838,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Erf-13
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L847"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L846"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_hardswish`
 
@@ -824,7 +860,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#hardswish-14
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L868"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L867"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_exp`
 
@@ -846,7 +882,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Exp-13
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L885"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L884"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_equal`
 
@@ -869,7 +905,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Equal-11
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L904"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L903"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_not`
 
@@ -891,7 +927,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Not-1
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L921"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L920"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_not_float`
 
@@ -913,7 +949,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Not-1
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L938"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L937"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_greater`
 
@@ -936,7 +972,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Greater-13
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L957"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L956"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_greater_float`
 
@@ -959,7 +995,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Greater-13
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L976"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L975"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_greater_or_equal`
 
@@ -982,7 +1018,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#GreaterOrEqual-12
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L995"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L994"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_greater_or_equal_float`
 
@@ -1005,7 +1041,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#GreaterOrEqual-12
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1014"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1013"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_less`
 
@@ -1028,7 +1064,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Less-13
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1033"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1032"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_less_float`
 
@@ -1051,7 +1087,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Less-13
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1052"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1051"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_less_or_equal`
 
@@ -1074,7 +1110,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#LessOrEqual-12
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1071"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1070"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_less_or_equal_float`
 
@@ -1097,7 +1133,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#LessOrEqual-12
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1090"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1089"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_identity`
 
@@ -1119,7 +1155,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Identity-14
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1128"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1127"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_transpose`
 
@@ -1142,7 +1178,48 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Transpose-13
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1206"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1142"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+
+## <kbd>function</kbd> `numpy_conv`
+
+```python
+numpy_conv(
+    x: ndarray,
+    w: ndarray,
+    b: Optional[ndarray] = None,
+    dilations: Tuple[int, ],
+    group: int = 1,
+    kernel_shape: Tuple[int, ],
+    pads: Tuple[int, ],
+    strides: Tuple[int, ]
+) → Tuple[ndarray]
+```
+
+Compute N-D convolution using Torch.
+
+Currently supports 2d convolution with torch semantics. This function is also ONNX compatible.
+
+See: https://github.com/onnx/onnx/blob/main/docs/Operators.md#Conv
+
+**Args:**
+
+- <b>`x`</b> (numpy.ndarray):  input data (many dtypes are supported). Shape is N x C x H x W for 2d
+- <b>`w`</b> (numpy.ndarray):  weights tensor. Shape is (O x I x Kh x Kw) for 2d
+- <b>`b`</b> (Optional\[numpy.ndarray\]):  bias tensor, Shape is (O,). Default to None.
+- <b>`dilations`</b> (Tuple\[int, ...\]):  dilation of the kernel, default 1 on all dimensions.
+- <b>`group`</b> (int):  number of convolution groups, can be 1 or a multiple of both (C,) and (O,), so  that I = C / group. Default to 1.
+- <b>`kernel_shape`</b> (Tuple\[int, ...\]):  shape of the kernel. Should have 2 elements for 2d conv
+- <b>`pads`</b> (Tuple\[int, ...\]):  padding in ONNX format (begin, end) on each axis
+- <b>`strides`</b> (Tuple\[int, ...\]):  stride of the convolution on each axis
+
+**Returns:**
+
+- <b>`res`</b> (numpy.ndarray):  a tensor of size (N x OutChannels x OutHeight x OutWidth).
+- <b>`See https`</b>: //pytorch.org/docs/stable/generated/torch.nn.Conv2d.html
+
+______________________________________________________________________
+
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1204"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_avgpool`
 
@@ -1181,7 +1258,7 @@ See: https://github.com/onnx/onnx/blob/main/docs/Operators.md#AveragePool
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1273"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1271"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_maxpool`
 
@@ -1222,7 +1299,7 @@ See: https://github.com/onnx/onnx/blob/main/docs/Operators.md#MaxPool
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1376"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1374"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_cast`
 
@@ -1247,7 +1324,7 @@ See: https://github.com/onnx/onnx/blob/main/docs/Operators.md#Cast
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1401"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1399"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_batchnorm`
 
@@ -1289,7 +1366,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#BatchNormalization-
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1472"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1470"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_flatten`
 
@@ -1312,7 +1389,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Flatten-13.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1490"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1488"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_or`
 
@@ -1335,7 +1412,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Or-7
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1508"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1506"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_or_float`
 
@@ -1358,7 +1435,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Or-7
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1526"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1524"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_round`
 
@@ -1380,7 +1457,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Round-11 Remark tha
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1543"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1541"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_pow`
 
@@ -1403,7 +1480,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Pow-13
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1666"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1664"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_floor`
 
@@ -1425,7 +1502,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Floor-1
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1680"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1678"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_max`
 
@@ -1450,7 +1527,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Max-1
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1697"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1695"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_min`
 
@@ -1475,7 +1552,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Max-1
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1714"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1712"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_sign`
 
@@ -1497,7 +1574,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sign-9
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1728"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1726"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_neg`
 
@@ -1519,7 +1596,7 @@ See https://github.com/onnx/onnx/blob/main/docs/Changelog.md#Sign-9
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1742"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/onnx/ops_impl.py#L1740"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `numpy_concatenate`
 
diff --git a/docs/developer-guide/api/concrete.ml.pytest.torch_models.md b/docs/developer-guide/api/concrete.ml.pytest.torch_models.md
index 6fca69e23..88d054e33 100644
--- a/docs/developer-guide/api/concrete.ml.pytest.torch_models.md
+++ b/docs/developer-guide/api/concrete.ml.pytest.torch_models.md
@@ -8,13 +8,13 @@ Torch modules for our pytests.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L16"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L18"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `SimpleNet`
 
 Fake torch model used to generate some onnx.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L19"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L21"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -24,7 +24,7 @@ __init__() → None
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L24"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L26"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -44,13 +44,13 @@ Forward function.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L38"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L40"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `FCSmall`
 
 Torch model for the tests.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L41"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L43"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -60,7 +60,7 @@ __init__(input_output, activation_function)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L47"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L49"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -79,13 +79,13 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L63"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L65"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `FC`
 
 Torch model for the tests.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L66"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L68"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -95,7 +95,7 @@ __init__(activation_function, input_output=3072)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L78"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L80"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -114,13 +114,13 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L100"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L102"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `CNN`
 
 Torch CNN model for the tests.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L103"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L105"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -130,7 +130,7 @@ __init__(input_output, activation_function)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L113"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L115"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -149,13 +149,13 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L131"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L133"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `CNNMaxPool`
 
 Torch CNN model for the tests with a max pool.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L134"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L136"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -165,7 +165,7 @@ __init__(input_output, activation_function)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L144"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L146"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -184,13 +184,13 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L162"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L164"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `CNNOther`
 
 Torch CNN model for the tests.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L165"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L167"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -200,7 +200,7 @@ __init__(input_output, activation_function)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L176"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L178"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -219,13 +219,13 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L195"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L197"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `CNNInvalid`
 
 Torch CNN model for the tests.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L198"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L200"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -235,7 +235,7 @@ __init__(activation_function, groups)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L214"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L216"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -254,13 +254,13 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L236"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L238"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `CNNGrouped`
 
 Torch CNN model with grouped convolution for compile torch tests.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L239"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L241"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -270,7 +270,7 @@ __init__(input_output, activation_function, groups)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L250"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L252"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -289,7 +289,7 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L268"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L270"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `NetWithLoops`
 
@@ -297,7 +297,7 @@ Torch model, where we reuse some elements in a loop.
 
 Torch model, where we reuse some elements in a loop in the forward and don't expect the user to define these elements in a particular order.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L275"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L277"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -307,7 +307,7 @@ __init__(activation_function, input_output, n_fc_layers)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L284"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L286"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -326,13 +326,13 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L301"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L303"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `MultiInputNN`
 
 Torch model to test multiple inputs forward.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L304"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L306"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -342,7 +342,7 @@ __init__(input_output, activation_function)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L308"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L310"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -362,13 +362,13 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L321"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L323"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `MultiInputNNConfigurable`
 
 Torch model to test multiple inputs forward.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L327"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L329"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -378,7 +378,7 @@ __init__(use_conv, use_qat, input_output, n_bits)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L337"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L339"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -398,13 +398,13 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L352"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L354"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `MultiInputNNDifferentSize`
 
 Torch model to test multiple inputs with different shape in the forward pass.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L355"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L357"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -419,7 +419,7 @@ __init__(
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L387"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L389"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -439,13 +439,13 @@ The output of the NN.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L407"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L409"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `BranchingModule`
 
 Torch model with some branching and skip connections.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L411"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L413"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -455,7 +455,7 @@ __init__(input_output, activation_function)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L416"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L418"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -474,13 +474,13 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L428"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L430"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `BranchingGemmModule`
 
 Torch model with some branching and skip connections.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L431"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L433"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -490,7 +490,7 @@ __init__(input_output, activation_function)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L437"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L439"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -509,13 +509,13 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L449"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L451"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `UnivariateModule`
 
 Torch model that calls univariate and shape functions of torch.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L453"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L455"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -525,7 +525,7 @@ __init__(input_output, activation_function)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L458"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L460"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -544,13 +544,13 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L474"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L476"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `StepActivationModule`
 
 Torch model implements a step function that needs Greater, Cast and Where.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L478"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L480"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -560,7 +560,7 @@ __init__(input_output, activation_function)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L483"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L485"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -579,13 +579,13 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L514"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L516"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `NetWithConcatUnsqueeze`
 
 Torch model to test the concat and unsqueeze operators.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L517"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L519"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -595,7 +595,7 @@ __init__(activation_function, input_output, n_fc_layers)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L526"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L528"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -614,13 +614,13 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L545"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L547"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `MultiOpOnSingleInputConvNN`
 
 Network that applies two quantized operations on a single input.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L548"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L550"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -630,7 +630,7 @@ __init__(can_remove_input_tlu: bool)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L554"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L556"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -649,7 +649,7 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L574"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L576"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `FCSeq`
 
@@ -657,7 +657,7 @@ Torch model that should generate MatMul->Add ONNX patterns.
 
 This network generates additions with a constant scalar
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L580"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L582"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -667,7 +667,7 @@ __init__(input_output, act)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L598"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L600"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -686,7 +686,7 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L612"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L614"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `FCSeqAddBiasVec`
 
@@ -694,7 +694,7 @@ Torch model that should generate MatMul->Add ONNX patterns.
 
 This network tests the addition with a constant vector
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L618"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L620"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -704,7 +704,7 @@ __init__(input_output, act)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L636"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L638"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -723,13 +723,13 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L650"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L652"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `TinyCNN`
 
 A very small CNN.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L653"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L655"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -746,7 +746,7 @@ Create the tiny CNN with two conv layers.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L668"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L670"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -765,7 +765,7 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L682"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L684"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `TinyQATCNN`
 
@@ -773,12 +773,19 @@ A very small QAT CNN to classify the sklearn digits data-set.
 
 This class also allows pruning to a maximum of 10 active neurons, which should help keep the accumulator bit-width low.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L689"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L691"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
 ```python
-__init__(n_classes, n_bits, n_active, signed, narrow) → None
+__init__(
+    n_classes,
+    n_bits,
+    n_active,
+    signed,
+    narrow,
+    power_of_two_scaling
+) → None
 ```
 
 Construct the CNN with a configurable number of classes.
@@ -790,10 +797,11 @@ Construct the CNN with a configurable number of classes.
 - <b>`n_active`</b> (int):  number of active (non-zero weight) neurons to keep
 - <b>`signed`</b> (bool):  whether quantized integer values are signed
 - <b>`narrow`</b> (bool):  whether the range of quantized integer values is narrow/symmetric
+- <b>`power_of_two_scaling`</b> (bool):  whether to use power-of-two scaling quantizers
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L759"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L801"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -812,27 +820,7 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L784"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
-
-### <kbd>method</kbd> `test_torch`
-
-```python
-test_torch(test_loader)
-```
-
-Test the network: measure accuracy on the test set.
-
-**Args:**
-
-- <b>`test_loader`</b>:  the test loader
-
-**Returns:**
-
-- <b>`res`</b>:  the number of correctly classified test examples
-
-______________________________________________________________________
-
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L727"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L769"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `toggle_pruning`
 
@@ -848,13 +836,13 @@ Enable or remove pruning.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L826"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L827"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `SimpleQAT`
 
 Torch model implements a step function that needs Greater, Cast and Where.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L829"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L830"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -864,7 +852,7 @@ __init__(input_output, activation_function, n_bits=2, disable_bit_check=False)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L865"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L866"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -883,13 +871,13 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L909"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L910"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QATTestModule`
 
 Torch model that implements a simple non-uniform quantizer.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L912"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L913"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -899,7 +887,7 @@ __init__(activation_function)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L917"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L918"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -918,13 +906,13 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L947"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L948"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `SingleMixNet`
 
 Torch model that with a single conv layer that produces the output, e.g., a blur filter.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L952"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L953"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -934,7 +922,7 @@ __init__(use_conv, use_qat, inp_size, n_bits)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L984"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L985"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -953,7 +941,7 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L997"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L998"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `DoubleQuantQATMixNet`
 
@@ -961,7 +949,7 @@ Torch model that with two different quantizers on the input.
 
 Used to test that it keeps the input TLU.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1003"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1004"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -971,7 +959,7 @@ __init__(use_conv, use_qat, inp_size, n_bits)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1021"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1022"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -990,13 +978,13 @@ the output of the NN
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1036"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1037"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `TorchSum`
 
 Torch model to test the ReduceSum ONNX operator in a leveled circuit.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1039"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1040"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -1013,7 +1001,7 @@ Initialize the module.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1055"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1056"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -1033,13 +1021,13 @@ Forward pass.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1068"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1069"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `TorchSumMod`
 
 Torch model to test the ReduceSum ONNX operator in a circuit containing a PBS.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1039"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1040"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -1056,7 +1044,7 @@ Initialize the module.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1071"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1072"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -1076,13 +1064,13 @@ Forward pass.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1088"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1089"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `NetWithConstantsFoldedBeforeOps`
 
 Torch QAT model that does not quantize the inputs.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1091"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1092"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -1097,7 +1085,7 @@ __init__(
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1118"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1119"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -1117,13 +1105,13 @@ Forward pass.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1137"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1138"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `ShapeOperationsNet`
 
 Torch QAT model that reshapes the input.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1140"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1141"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -1133,7 +1121,7 @@ __init__(is_qat)
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1146"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1147"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -1153,13 +1141,13 @@ Forward pass.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1186"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1187"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `PaddingNet`
 
 Torch QAT model that applies various padding patterns.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1189"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1190"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -1169,7 +1157,7 @@ __init__()
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1195"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1196"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -1189,13 +1177,13 @@ Forward pass.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1222"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1223"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantCustomModel`
 
 A small quantized network with Brevitas, trained on make_classification.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1225"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1226"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -1206,7 +1194,8 @@ __init__(
     hidden_shape: int = 100,
     n_bits: int = 5,
     act_quant=<class 'brevitas.quant.scaled_int.Int8ActPerTensorFloat'>,
-    weight_quant=<class 'brevitas.quant.scaled_int.Int8WeightPerTensorFloat'>
+    weight_quant=<class 'brevitas.quant.scaled_int.Int8WeightPerTensorFloat'>,
+    bias_quant=None
 )
 ```
 
@@ -1220,10 +1209,11 @@ Quantized Torch Model with Brevitas.
 - <b>`n_bits`</b> (int):  Bit of quantization
 - <b>`weight_quant`</b> (brevitas.quant):  Quantization protocol of weights
 - <b>`act_quant`</b> (brevitas.quant):  Quantization protocol of activations.
+- <b>`bias_quant`</b> (brevitas.quant):  Quantizer for the linear layer bias
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1280"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1285"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -1243,13 +1233,13 @@ Forward pass.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1298"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1303"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `TorchCustomModel`
 
 A small network with Brevitas, trained on make_classification.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1301"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1306"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -1267,7 +1257,7 @@ Torch Model.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1314"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1319"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -1287,13 +1277,13 @@ Forward pass.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1329"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1334"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `ConcatFancyIndexing`
 
 Concat with fancy indexing.
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1332"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1337"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -1319,7 +1309,7 @@ Torch Model.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/pytest/torch_models.py#L1360"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/pytest/torch_models.py#L1365"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
diff --git a/docs/developer-guide/api/concrete.ml.quantization.md b/docs/developer-guide/api/concrete.ml.quantization.md
index 33a2d5359..8fb4af6c4 100644
--- a/docs/developer-guide/api/concrete.ml.quantization.md
+++ b/docs/developer-guide/api/concrete.ml.quantization.md
@@ -12,4 +12,6 @@ Modules for quantization.
 - **base_quantized_op**
 - **quantized_module**
 - **quantized_ops**
+- **quantized_module_passes**
 - **post_training**
+- **qat_quantizers**
diff --git a/docs/developer-guide/api/concrete.ml.quantization.post_training.md b/docs/developer-guide/api/concrete.ml.quantization.post_training.md
index 510187483..0ac4b03c9 100644
--- a/docs/developer-guide/api/concrete.ml.quantization.post_training.md
+++ b/docs/developer-guide/api/concrete.ml.quantization.post_training.md
@@ -14,7 +14,7 @@ Post Training Quantization methods.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/post_training.py#L26"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/post_training.py#L27"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>function</kbd> `get_n_bits_dict`
 
@@ -36,7 +36,7 @@ Convert the n_bits parameter into a proper dictionary.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/post_training.py#L92"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/post_training.py#L93"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `ONNXConverter`
 
@@ -54,7 +54,7 @@ This class should be sub-classed to provide specific calibration and quantizatio
 - <b>`numpy_model`</b> (NumpyModule):  Model in numpy.
 - <b>`rounding_threshold_bits`</b> (int):  if not None, every accumulators in the model are rounded down  to the given bits of precision
 
-<a href="../../../src/concrete/ml/quantization/post_training.py#L128"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/post_training.py#L129"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -108,7 +108,7 @@ Get the number of bits to use for the quantization of any constants (usually wei
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/post_training.py#L577"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/post_training.py#L580"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `quantize_module`
 
@@ -130,7 +130,7 @@ Following https://arxiv.org/abs/1712.05877 guidelines.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/post_training.py#L719"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/post_training.py#L725"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `PostTrainingAffineQuantization`
 
@@ -153,7 +153,7 @@ Create the quantized version of the passed numpy module.
 
 - <b>`QuantizedModule`</b>:  A quantized version of the numpy model.
 
-<a href="../../../src/concrete/ml/quantization/post_training.py#L128"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/post_training.py#L129"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -207,7 +207,7 @@ Get the number of bits to use for the quantization of any constants (usually wei
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/post_training.py#L577"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/post_training.py#L580"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `quantize_module`
 
@@ -229,7 +229,7 @@ Following https://arxiv.org/abs/1712.05877 guidelines.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/post_training.py#L863"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/post_training.py#L869"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `PostTrainingQATImporter`
 
@@ -237,7 +237,7 @@ Converter of Quantization Aware Training networks.
 
 This class provides specific configuration for QAT networks during ONNX network conversion to Concrete ML computation graphs.
 
-<a href="../../../src/concrete/ml/quantization/post_training.py#L128"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/post_training.py#L129"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -291,7 +291,7 @@ Get the number of bits to use for the quantization of any constants (usually wei
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/post_training.py#L577"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/post_training.py#L580"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `quantize_module`
 
diff --git a/docs/developer-guide/api/concrete.ml.quantization.quantized_module_passes.md b/docs/developer-guide/api/concrete.ml.quantization.quantized_module_passes.md
new file mode 100644
index 000000000..56c78602a
--- /dev/null
+++ b/docs/developer-guide/api/concrete.ml.quantization.quantized_module_passes.md
@@ -0,0 +1,143 @@
+<!-- markdownlint-disable -->
+
+<a href="../../../src/concrete/ml/quantization/quantized_module_passes.py#L0"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+
+# <kbd>module</kbd> `concrete.ml.quantization.quantized_module_passes`
+
+Optimization passes for QuantizedModules.
+
+______________________________________________________________________
+
+<a href="../../../src/concrete/ml/quantization/quantized_module_passes.py#L29"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+
+## <kbd>class</kbd> `PowerOfTwoScalingRoundPBSAdapter`
+
+Detect neural network patterns that can be optimized with round PBS.
+
+<a href="../../../src/concrete/ml/quantization/quantized_module_passes.py#L39"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+
+### <kbd>method</kbd> `__init__`
+
+```python
+__init__(qmodule: QuantizedModule) → None
+```
+
+______________________________________________________________________
+
+#### <kbd>property</kbd> num_ignored_valid_patterns
+
+Get the number of optimizable patterns that were ignored.
+
+Patterns could be ignored since a number of rounding bits was set manually through the compilation function.
+
+**Returns:**
+
+- <b>`result`</b> (int):  number of patterns that could be optimized but were not
+
+______________________________________________________________________
+
+<a href="../../../src/concrete/ml/quantization/quantized_module_passes.py#L89"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+
+### <kbd>method</kbd> `compute_op_predecessors`
+
+```python
+compute_op_predecessors() → DefaultDict[Union[QuantizedOp, NoneType], List[Tuple[Union[QuantizedOp, NoneType], str]]]
+```
+
+Compute the predecessors for each QuantizedOp in a QuantizedModule.
+
+Stores, for each quantized op, a list of quantized ops that produce its inputs. Currently only the first input of the operations is considered as it is, usually, the encrypted input.
+
+**Returns:**
+
+- <b>`result`</b> (PredecessorsType):  a dictionary containing a hierarchy of op  predecessors
+
+______________________________________________________________________
+
+<a href="../../../src/concrete/ml/quantization/quantized_module_passes.py#L151"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+
+### <kbd>method</kbd> `detect_patterns`
+
+```python
+detect_patterns(
+    predecessors: DefaultDict[Optional[QuantizedOp], List[Tuple[Optional[QuantizedOp], str]]]
+) → Dict[QuantizedMixingOp, Tuple[List[Union[QuantizedOp, NoneType]], Union[QuantizedOp, NoneType]]]
+```
+
+Detect the patterns that can be optimized with roundPBS in the QuantizedModule.
+
+**Args:**
+
+- <b>`predecessors`</b> (PredecessorsType):  Module predecessor operation list
+
+**Returns:**
+
+- <b>`result`</b> (PatternDict):  list of optimizable patterns
+
+______________________________________________________________________
+
+<a href="../../../src/concrete/ml/quantization/quantized_module_passes.py#L118"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+
+### <kbd>method</kbd> `match_path_pattern`
+
+```python
+match_path_pattern(
+    predecessors: DefaultDict[Optional[QuantizedOp], List[Tuple[Optional[QuantizedOp], str]]],
+    nodes_in_path: List[Optional[QuantizedOp]],
+    input_producer_of_path: Optional[QuantizedOp]
+) → bool
+```
+
+Determine if a pattern has the structure that makes it viable for roundPBS.
+
+**Args:**
+
+- <b>`predecessors`</b> (PredecessorsType):  Module predecessor operation list
+- <b>`nodes_in_path`</b> (List\[QuantizedOp\]):  list of quantized ops in the pattern
+- <b>`input_producer_of_path`</b> (Optional\[QuantizedOp\]):  operation that produces the input
+
+**Returns:**
+
+- <b>`result`</b> (bool):  whether the pattern can be optimized
+
+______________________________________________________________________
+
+<a href="../../../src/concrete/ml/quantization/quantized_module_passes.py#L55"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+
+### <kbd>method</kbd> `process`
+
+```python
+process() → Dict[QuantizedMixingOp, Tuple[List[Union[QuantizedOp, NoneType]], Union[QuantizedOp, NoneType]]]
+```
+
+Analyze an ONNX graph and detect Gemm/Conv patterns that can use RoundPBS.
+
+We want to detect a gemm/conv node whose weights/bias are Brevitas QAT, and whose input is produced by a Brevitas QAT node that is applied on the output of another Gemm/conv node. Optionally a Relu can be placed before this input quantization node.
+
+Nothing will be done if rounding is already specified.
+
+**Returns:**
+
+- <b>`result`</b> (PatternDict):  a dictionary containing for each Conv/Gemm node for which  round PBS can be applied based on power-of-two scaling factors
+
+______________________________________________________________________
+
+<a href="../../../src/concrete/ml/quantization/quantized_module_passes.py#L218"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+
+### <kbd>method</kbd> `process_patterns`
+
+```python
+process_patterns(
+    valid_paths: Dict[QuantizedMixingOp, Tuple[List[Optional[QuantizedOp]], Optional[QuantizedOp]]]
+) → Dict[QuantizedMixingOp, Tuple[List[Union[QuantizedOp, NoneType]], Union[QuantizedOp, NoneType]]]
+```
+
+Configure the rounding bits of roundPBS for the optimizable operations.
+
+**Args:**
+
+- <b>`valid_paths`</b> (PatternDict):  list of optimizable patterns
+
+**Returns:**
+
+- <b>`result`</b> (PatternDict):  list of patterns actually optimized with roundPBS
diff --git a/docs/developer-guide/api/concrete.ml.quantization.quantized_ops.md b/docs/developer-guide/api/concrete.ml.quantization.quantized_ops.md
index 288836e0c..e1b6d7166 100644
--- a/docs/developer-guide/api/concrete.ml.quantization.quantized_ops.md
+++ b/docs/developer-guide/api/concrete.ml.quantization.quantized_ops.md
@@ -259,7 +259,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L166"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L167"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
@@ -273,7 +273,7 @@ q_impl(
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L300"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L315"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedMatMul`
 
@@ -306,7 +306,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L166"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L167"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
@@ -320,7 +320,7 @@ q_impl(
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L306"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L321"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedAdd`
 
@@ -340,7 +340,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L412"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L427"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `can_fuse`
 
@@ -358,7 +358,7 @@ Add operation can be computed in float and fused if it operates over inputs prod
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L315"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L330"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
@@ -371,7 +371,7 @@ q_impl(
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L426"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L441"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedTanh`
 
@@ -389,7 +389,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L432"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L447"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedSoftplus`
 
@@ -407,7 +407,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L438"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L453"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedExp`
 
@@ -425,7 +425,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L444"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L459"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedLog`
 
@@ -443,7 +443,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L450"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L465"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedAbs`
 
@@ -461,7 +461,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L456"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L471"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedIdentity`
 
@@ -479,7 +479,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L461"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L476"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
@@ -492,7 +492,7 @@ q_impl(
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L475"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L490"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedReshape`
 
@@ -510,7 +510,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L516"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L531"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `can_fuse`
 
@@ -528,7 +528,7 @@ Max Pooling operation can not be fused since it must be performed over integer t
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L481"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L496"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
@@ -552,13 +552,13 @@ Reshape the input integer encrypted tensor.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L528"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L543"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedConv`
 
 Quantized Conv op.
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L533"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L548"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -601,7 +601,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L598"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L614"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
@@ -632,13 +632,13 @@ Allows an optional quantized bias.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L779"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L805"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedAvgPool`
 
 Quantized Average Pooling op.
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L785"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L811"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -665,7 +665,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L831"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L857"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
@@ -679,13 +679,13 @@ q_impl(
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L922"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L948"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedMaxPool`
 
 Quantized Max Pooling op.
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L928"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L954"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -712,7 +712,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1039"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1065"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `can_fuse`
 
@@ -730,7 +730,7 @@ Max Pooling operation can not be fused since it must be performed over integer t
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L976"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1002"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
@@ -743,13 +743,13 @@ q_impl(
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1051"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1077"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedPad`
 
 Quantized Padding op.
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1056"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1082"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -776,7 +776,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1123"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1149"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `can_fuse`
 
@@ -794,7 +794,7 @@ Pad operation cannot be fused since it must be performed over integer tensors.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1082"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1108"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
@@ -808,7 +808,7 @@ q_impl(
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1134"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1160"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedWhere`
 
@@ -816,7 +816,7 @@ Where operator on quantized arrays.
 
 Supports only constants for the results produced on the True/False branches.
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1143"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1169"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -843,7 +843,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1174"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1200"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedCast`
 
@@ -863,7 +863,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1183"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1209"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedGreater`
 
@@ -871,7 +871,7 @@ Comparison operator >.
 
 Only supports comparison with a constant.
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1192"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1218"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -898,7 +898,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1215"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1241"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedGreaterOrEqual`
 
@@ -906,7 +906,7 @@ Comparison operator >=.
 
 Only supports comparison with a constant.
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1224"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1250"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -933,7 +933,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1247"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1273"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedLess`
 
@@ -941,7 +941,7 @@ Comparison operator \<.
 
 Only supports comparison with a constant.
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1256"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1282"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -968,7 +968,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1279"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1305"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedLessOrEqual`
 
@@ -976,7 +976,7 @@ Comparison operator \<=.
 
 Only supports comparison with a constant.
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1288"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1314"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -1003,7 +1003,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1311"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1337"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedOr`
 
@@ -1023,7 +1023,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1321"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1347"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedDiv`
 
@@ -1043,7 +1043,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1331"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1357"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedMul`
 
@@ -1063,7 +1063,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1341"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1367"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedSub`
 
@@ -1083,7 +1083,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L412"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L427"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `can_fuse`
 
@@ -1101,7 +1101,7 @@ Add operation can be computed in float and fused if it operates over inputs prod
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L315"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L330"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
@@ -1114,7 +1114,7 @@ q_impl(
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1351"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1377"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedBatchNormalization`
 
@@ -1132,7 +1132,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1357"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1383"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedFlatten`
 
@@ -1150,7 +1150,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1404"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1430"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `can_fuse`
 
@@ -1168,7 +1168,7 @@ Flatten operation cannot be fused since it must be performed over integer tensor
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1363"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1389"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
@@ -1192,13 +1192,13 @@ Flatten the input integer encrypted tensor.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1416"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1442"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedReduceSum`
 
 ReduceSum with encrypted input.
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1421"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1447"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -1239,7 +1239,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1471"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1497"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `calibrate`
 
@@ -1259,7 +1259,7 @@ Create corresponding QuantizedArray for the output of the activation function.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1495"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1521"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
@@ -1283,7 +1283,7 @@ Sum the encrypted tensor's values along the given axes.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1618"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1644"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedErf`
 
@@ -1301,7 +1301,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1624"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1650"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedNot`
 
@@ -1319,13 +1319,13 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1630"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1656"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedBrevitasQuant`
 
 Brevitas uniform quantization with encrypted input.
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1640"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1666"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -1368,7 +1368,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1742"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1768"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `calibrate`
 
@@ -1388,7 +1388,7 @@ Create corresponding QuantizedArray for the output of Quantization function.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1768"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1794"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
@@ -1412,7 +1412,7 @@ Quantize values.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1833"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1859"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedTranspose`
 
@@ -1432,7 +1432,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1875"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1901"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `can_fuse`
 
@@ -1450,7 +1450,7 @@ Transpose can not be fused since it must be performed over integer tensors as it
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1843"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1869"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
@@ -1474,7 +1474,7 @@ Transpose the input integer encrypted tensor.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1887"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1913"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedFloor`
 
@@ -1492,7 +1492,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1893"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1919"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedMax`
 
@@ -1510,7 +1510,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1899"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1925"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedMin`
 
@@ -1528,7 +1528,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1905"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1931"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedNeg`
 
@@ -1546,7 +1546,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1911"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1937"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedSign`
 
@@ -1564,7 +1564,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1917"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1943"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedUnsqueeze`
 
@@ -1582,7 +1582,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1957"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1983"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `can_fuse`
 
@@ -1600,7 +1600,7 @@ Unsqueeze can not be fused since it must be performed over integer tensors as it
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1923"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1949"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
@@ -1624,7 +1624,7 @@ Unsqueeze the input tensors on a given axis.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1969"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1995"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedConcat`
 
@@ -1642,7 +1642,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2025"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2051"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `can_fuse`
 
@@ -1660,7 +1660,7 @@ Concatenation can not be fused since it must be performed over integer tensors a
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L1975"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2001"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
@@ -1684,7 +1684,7 @@ Concatenate the input tensors on a given axis.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2037"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2063"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `QuantizedSqueeze`
 
@@ -1702,7 +1702,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2077"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2103"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `can_fuse`
 
@@ -1720,7 +1720,7 @@ Squeeze can not be fused since it must be performed over integer tensors as it r
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2043"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2069"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
@@ -1744,7 +1744,7 @@ Squeeze the input tensors on a given axis.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2089"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2115"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `ONNXShape`
 
@@ -1762,7 +1762,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2104"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2130"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `can_fuse`
 
@@ -1780,7 +1780,7 @@ This operation returns the shape of the tensor and thus can not be fused into a
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2094"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2120"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
@@ -1793,7 +1793,7 @@ q_impl(
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2116"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2142"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `ONNXConstantOfShape`
 
@@ -1811,7 +1811,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2121"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2147"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `can_fuse`
 
@@ -1829,7 +1829,7 @@ This operation returns a new encrypted tensor and thus can not be fused.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2133"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2159"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `ONNXGather`
 
@@ -1849,7 +1849,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2165"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2191"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `can_fuse`
 
@@ -1867,7 +1867,7 @@ This operation returns values from a tensor and thus can not be fused into a uni
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2141"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2167"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
@@ -1880,7 +1880,7 @@ q_impl(
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2177"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2203"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `ONNXSlice`
 
@@ -1898,7 +1898,7 @@ Get the names of encrypted integer tensors that are used by this op.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2206"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2232"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `can_fuse`
 
@@ -1916,7 +1916,7 @@ This operation returns values from a tensor and thus can not be fused into a uni
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2182"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/quantization/quantized_ops.py#L2208"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `q_impl`
 
diff --git a/docs/developer-guide/api/concrete.ml.sklearn.qnn_module.md b/docs/developer-guide/api/concrete.ml.sklearn.qnn_module.md
index b4ffca6ff..efbe95b98 100644
--- a/docs/developer-guide/api/concrete.ml.sklearn.qnn_module.md
+++ b/docs/developer-guide/api/concrete.ml.sklearn.qnn_module.md
@@ -12,7 +12,7 @@ Sparse Quantized Neural Network torch module.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/sklearn/qnn_module.py#L14"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/sklearn/qnn_module.py#L16"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ## <kbd>class</kbd> `SparseQuantNeuralNetwork`
 
@@ -20,7 +20,7 @@ Sparse Quantized Neural Network.
 
 This class implements an MLP that is compatible with FHE constraints. The weights and activations are quantized to low bit-width and pruning is used to ensure accumulators do not surpass an user-provided accumulator bit-width. The number of classes and number of layers are specified by the user, as well as the breadth of the network
 
-<a href="../../../src/concrete/ml/sklearn/qnn_module.py#L24"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/sklearn/qnn_module.py#L26"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `__init__`
 
@@ -36,7 +36,8 @@ __init__(
     n_prune_neurons_percentage: float = 0.0,
     activation_function: Type = <class 'torch.nn.modules.activation.ReLU'>,
     quant_narrow: bool = False,
-    quant_signed: bool = True
+    quant_signed: bool = True,
+    power_of_two_scaling: bool = True
 )
 ```
 
@@ -55,6 +56,7 @@ Sparse Quantized Neural Network constructor.
 - <b>`activation_function`</b> (Type):  The activation function to use in the network  (e.g., torch.ReLU, torch.SELU, torch.Sigmoid, ...).
 - <b>`quant_narrow`</b> (bool):  Whether this network should quantize the values using narrow range  (e.g a 2-bits signed quantization uses \[-1, 0, 1\] instead of \[-2, -1, 0, 1\]).
 - <b>`quant_signed`</b> (bool):  Whether this network should quantize the values using signed  integers.
+- <b>`power_of_two_scaling`</b> (bool):  Force quantization scales to be a power of two  to enable inference speed optimizations. Defaults to True
 
 **Raises:**
 
@@ -62,7 +64,7 @@ Sparse Quantized Neural Network constructor.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/sklearn/qnn_module.py#L228"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/sklearn/qnn_module.py#L237"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `enable_pruning`
 
@@ -78,7 +80,7 @@ Enable pruning in the network. Pruning must be made permanent to recover pruned
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/sklearn/qnn_module.py#L284"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/sklearn/qnn_module.py#L293"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `forward`
 
@@ -98,7 +100,7 @@ Forward pass.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/sklearn/qnn_module.py#L158"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/sklearn/qnn_module.py#L167"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `make_pruning_permanent`
 
@@ -110,7 +112,7 @@ Make the learned pruning permanent in the network.
 
 ______________________________________________________________________
 
-<a href="../../../src/concrete/ml/sklearn/qnn_module.py#L138"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
+<a href="../../../src/concrete/ml/sklearn/qnn_module.py#L147"><img align="right" style="float:right;" src="https://img.shields.io/badge/-source-cccccc?style=flat-square"></a>
 
 ### <kbd>method</kbd> `max_active_neurons`
 
diff --git a/script/doc_utils/update_apidocs.sh b/script/doc_utils/update_apidocs.sh
index 42d207a37..056ff7f2d 100755
--- a/script/doc_utils/update_apidocs.sh
+++ b/script/doc_utils/update_apidocs.sh
@@ -7,9 +7,12 @@ APIDOCS_OUTPUT="$1"
 # Clean
 rm -rf "$APIDOCS_OUTPUT"
 
+# Ignore concrete.ml.quantization.qat_quantizers since 
+# brevitas has some issues with lazydocs
 poetry run lazydocs --output-path="$APIDOCS_OUTPUT" \
     --overview-file="README.md" \
     --src-base-url="../../../" \
     --no-watermark \
+    --ignored-modules concrete.ml.quantization.qat_quantizers \
     concrete.ml
 
diff --git a/src/concrete/ml/common/utils.py b/src/concrete/ml/common/utils.py
index 3af1800b5..4414056fb 100644
--- a/src/concrete/ml/common/utils.py
+++ b/src/concrete/ml/common/utils.py
@@ -42,6 +42,14 @@
 # Indicate if the old simulation method should be used when simulating FHE executions
 USE_OLD_VL = True
 
+# Debug option for testing round PBS optimization
+# Setting this option to true will make quantizers "round half up"
+# For example: 0.5 -> 1, 1.5 -> 2 instead of "round half to even"
+# When the option is set to false, Concrete ML uses numpy.rint
+# which has the same behavior as torch.round -> Brevitas nets
+# should be exact compared to their Concrete ML QuantizedModule
+QUANT_ROUND_LIKE_ROUND_PBS = False
+
 
 class FheMode(str, enum.Enum):
     """Enum representing the execution mode.
diff --git a/src/concrete/ml/onnx/ops_impl.py b/src/concrete/ml/onnx/ops_impl.py
index 4b08818ce..d06489a49 100644
--- a/src/concrete/ml/onnx/ops_impl.py
+++ b/src/concrete/ml/onnx/ops_impl.py
@@ -14,6 +14,7 @@
 from scipy import special
 from typing_extensions import SupportsIndex
 
+from ..common import utils
 from ..common.debugging import assert_false, assert_true
 from .onnx_impl_utils import (
     compute_onnx_pool_padding,
@@ -240,7 +241,6 @@ def numpy_constant(**kwargs):
 # pylint: disable=invalid-name
 # 1 is technically an int but is accepted by mypy as a float (and it simplifies our life for
 # compilation) so instead of passing 1.0 by default 1 is passed
-@onnx_func_raw_args("c")
 def numpy_gemm(
     a: numpy.ndarray,
     b: numpy.ndarray,
@@ -1140,7 +1140,6 @@ def numpy_transpose(x: numpy.ndarray, *, perm=None) -> Tuple[numpy.ndarray]:
     return (numpy.transpose(x, axes=perm),)
 
 
-@onnx_func_raw_args("b")
 def numpy_conv(
     x: numpy.ndarray,
     w: numpy.ndarray,
@@ -1655,7 +1654,10 @@ def numpy_brevitas_quant(
     y = numpy.clip(y, min_int_val, max_int_val)
 
     # Quantize to produce integers representing the float quantized values
-    y = numpy.rint(y)
+    if utils.QUANT_ROUND_LIKE_ROUND_PBS:
+        y = numpy.floor(y + 0.5)
+    else:
+        y = numpy.rint(y)
 
     # Compute quantized floating point values
     y = (y - zero_point) * scale
diff --git a/src/concrete/ml/pytest/torch_models.py b/src/concrete/ml/pytest/torch_models.py
index b1a79720e..bf371c7fd 100644
--- a/src/concrete/ml/pytest/torch_models.py
+++ b/src/concrete/ml/pytest/torch_models.py
@@ -6,10 +6,12 @@
 import brevitas.nn as qnn
 import numpy
 import torch
-from brevitas.quant import Int8ActPerTensorFloat, Int8WeightPerTensorFloat
+from brevitas.quant import Int8ActPerTensorFloat, Int8WeightPerTensorFloat, IntBias
 from torch import nn
 from torch.nn.utils import prune
 
+from concrete.ml.quantization.qat_quantizers import Int8ActPerTensorPoT, Int8WeightPerTensorPoT
+
 # pylint: disable=too-many-lines
 
 
@@ -686,7 +688,7 @@ class TinyQATCNN(nn.Module):
     should help keep the accumulator bit-width low.
     """
 
-    def __init__(self, n_classes, n_bits, n_active, signed, narrow) -> None:
+    def __init__(self, n_classes, n_bits, n_active, signed, narrow, power_of_two_scaling) -> None:
         """Construct the CNN with a configurable number of classes.
 
         Args:
@@ -695,6 +697,8 @@ def __init__(self, n_classes, n_bits, n_active, signed, narrow) -> None:
             n_active (int): number of active (non-zero weight) neurons to keep
             signed (bool): whether quantized integer values are signed
             narrow (bool): whether the range of quantized integer values is narrow/symmetric
+            power_of_two_scaling (bool): whether to use power-of-two scaling quantizers which
+                allows to test the round PBS optimization when the scales are power-of-two
         """
         super().__init__()
 
@@ -705,17 +709,56 @@ def __init__(self, n_classes, n_bits, n_active, signed, narrow) -> None:
 
         q_args = {"signed": signed, "narrow_range": narrow}
 
-        self.quant1 = qnn.QuantIdentity(bit_width=a_bits, return_quant_tensor=True, **q_args)
+        if power_of_two_scaling:
+            act_quant = Int8ActPerTensorPoT
+            weight_quant = Int8WeightPerTensorPoT
+            bias_quant = IntBias
+        else:
+            act_quant = Int8ActPerTensorFloat
+            weight_quant = Int8WeightPerTensorFloat
+            bias_quant = None
+
+        self.quant1 = qnn.QuantIdentity(
+            bit_width=a_bits, return_quant_tensor=True, **q_args, act_quant=act_quant
+        )
         self.conv1 = qnn.QuantConv2d(
-            1, 2, 3, stride=1, padding=0, weight_bit_width=w_bits, **q_args
+            1,
+            2,
+            3,
+            stride=1,
+            padding=0,
+            weight_bit_width=w_bits,
+            **q_args,
+            weight_quant=weight_quant,
+            bias_quant=bias_quant,
+        )
+        self.quant2 = qnn.QuantIdentity(
+            bit_width=a_bits, return_quant_tensor=True, **q_args, act_quant=act_quant
         )
-        self.quant2 = qnn.QuantIdentity(bit_width=a_bits, return_quant_tensor=True, **q_args)
         self.conv2 = qnn.QuantConv2d(
-            2, 3, 3, stride=2, padding=0, weight_bit_width=w_bits, **q_args
+            2,
+            3,
+            3,
+            stride=2,
+            padding=0,
+            weight_bit_width=w_bits,
+            **q_args,
+            weight_quant=weight_quant,
+            bias_quant=bias_quant,
+        )
+        self.quant3 = qnn.QuantIdentity(
+            bit_width=a_bits, return_quant_tensor=True, **q_args, act_quant=act_quant
         )
-        self.quant3 = qnn.QuantIdentity(bit_width=a_bits, return_quant_tensor=True, **q_args)
         self.conv3 = qnn.QuantConv2d(
-            3, 16, 2, stride=1, padding=0, weight_bit_width=w_bits, **q_args
+            3,
+            16,
+            2,
+            stride=1,
+            padding=0,
+            weight_bit_width=w_bits,
+            **q_args,
+            weight_quant=weight_quant,
+            bias_quant=bias_quant,
         )
 
         self.quant4 = qnn.QuantIdentity(bit_width=a_bits, return_quant_tensor=True, **q_args)
@@ -781,47 +824,6 @@ def forward(self, x):
         x = self.fc1(x)
         return x
 
-    def test_torch(self, test_loader):
-        """Test the network: measure accuracy on the test set.
-
-        Args:
-            test_loader: the test loader
-
-        Returns:
-            res: the number of correctly classified test examples
-
-        """
-
-        # Freeze normalization layers
-        self.eval()
-
-        all_y_pred = numpy.zeros((len(test_loader)), dtype=numpy.int64)
-        all_targets = numpy.zeros((len(test_loader)), dtype=numpy.int64)
-
-        # Iterate over the batches
-        idx = 0
-        for data, target in test_loader:
-            # Accumulate the ground truth labels
-            endidx = idx + target.shape[0]
-            all_targets[idx:endidx] = target.numpy()
-
-            # Run forward and get the raw predictions first
-            raw_pred = self(data).detach().numpy()
-
-            # Get the predicted class id, handle NaNs
-            if numpy.any(numpy.isnan(raw_pred)):
-                output = -1  # pragma: no cover
-            else:
-                output = raw_pred.argmax(1)
-
-            all_y_pred[idx:endidx] = output
-
-            idx += target.shape[0]
-
-        # Print out the accuracy as a percentage
-        n_correct = numpy.sum(all_targets == all_y_pred)
-        return n_correct
-
 
 class SimpleQAT(nn.Module):
     """Torch model implements a step function that needs Greater, Cast and Where."""
@@ -1230,6 +1232,7 @@ def __init__(
         n_bits: int = 5,
         act_quant=Int8ActPerTensorFloat,
         weight_quant=Int8WeightPerTensorFloat,
+        bias_quant=None,
     ):
         """Quantized Torch Model with Brevitas.
 
@@ -1240,7 +1243,7 @@ def __init__(
             n_bits (int): Bit of quantization
             weight_quant (brevitas.quant): Quantization protocol of weights
             act_quant (brevitas.quant): Quantization protocol of activations.
-
+            bias_quant (brevitas.quant): Quantizer for the linear layer bias
         """
         super().__init__()
 
@@ -1253,6 +1256,7 @@ def __init__(
             weight_bit_width=n_bits,
             weight_quant=weight_quant,
             bias=True,
+            bias_quant=bias_quant,
             return_quant_tensor=True,
         )
 
@@ -1263,6 +1267,7 @@ def __init__(
             weight_bit_width=n_bits,
             weight_quant=weight_quant,
             bias=True,
+            bias_quant=bias_quant,
             return_quant_tensor=True,
         )
 
@@ -1274,6 +1279,7 @@ def __init__(
             weight_bit_width=n_bits,
             weight_quant=weight_quant,
             bias=True,
+            bias_quant=bias_quant,
             return_quant_tensor=True,
         )
 
diff --git a/src/concrete/ml/pytest/utils.py b/src/concrete/ml/pytest/utils.py
index a66bb448e..ce130991c 100644
--- a/src/concrete/ml/pytest/utils.py
+++ b/src/concrete/ml/pytest/utils.py
@@ -173,6 +173,9 @@ def instantiate_model_generic(model_class, n_bits, **parameters):
             extra_kwargs["module__n_a_bits"] = 2
             extra_kwargs["module__n_accum_bits"] = 7
 
+        # Disable power-of-two since it sets the input bitwidth to 8
+        # and thus increases bitwidth too much for a test
+        extra_kwargs["module__power_of_two_scaling"] = False
         extra_kwargs.update(parameters)
         model = model_class(**extra_kwargs)
 
diff --git a/src/concrete/ml/quantization/__init__.py b/src/concrete/ml/quantization/__init__.py
index 4e6dc0594..845b5dc11 100644
--- a/src/concrete/ml/quantization/__init__.py
+++ b/src/concrete/ml/quantization/__init__.py
@@ -7,6 +7,7 @@
     QuantizedAdd,
     QuantizedAvgPool,
     QuantizedBatchNormalization,
+    QuantizedBrevitasQuant,
     QuantizedCelu,
     QuantizedClip,
     QuantizedConv,
diff --git a/src/concrete/ml/quantization/post_training.py b/src/concrete/ml/quantization/post_training.py
index 2a5edda20..fcbe6abd5 100644
--- a/src/concrete/ml/quantization/post_training.py
+++ b/src/concrete/ml/quantization/post_training.py
@@ -19,6 +19,7 @@
     QuantizedOp,
 )
 from .quantized_module import QuantizedModule
+from .quantized_module_passes import PowerOfTwoScalingRoundPBSAdapter
 from .quantized_ops import QuantizedBrevitasQuant
 from .quantizers import QuantizationOptions, QuantizedArray, UniformQuantizer
 
@@ -427,13 +428,14 @@ def _quantize_layers(self, *input_calibration_data: numpy.ndarray):
                 attributes.update({"rounding_threshold_bits": self.rounding_threshold_bits})
 
             # All inputs, allow optional constants (they become None)
-            curr_inputs = {
-                input_name: node_results.get(input_name, None) for input_name in node.input
-            }
+            # Note that input of a node can be duplicated, e.g., (%a, %a, %b)
+            curr_inputs = [
+                (input_name, node_results.get(input_name, None)) for input_name in node.input
+            ]
 
             # Constant inputs
             curr_cst_inputs: Dict[int, ONNXOpInputOutputType] = {}
-            for input_idx, (input_name, value) in enumerate(curr_inputs.items()):
+            for input_idx, (input_name, value) in enumerate(curr_inputs):
                 if not (input_name in self.quant_params or input_name in constants):
                     continue
 
@@ -455,10 +457,12 @@ def _quantize_layers(self, *input_calibration_data: numpy.ndarray):
             has_variable_inputs = (len(curr_inputs) - len(curr_cst_inputs)) > 0
 
             variable_input_names = [
-                input_name for input_name in curr_inputs if input_name not in constants
+                input_name for input_name, _ in curr_inputs if input_name not in constants
             ]
             curr_calibration_data = tuple(
-                curr_inputs[input_name] for input_name in variable_input_names
+                input_data
+                for input_name, input_data in curr_inputs
+                if input_name in variable_input_names
             )
 
             # For mypy
@@ -604,6 +608,10 @@ def quantize_module(self, *calibration_data: numpy.ndarray) -> QuantizedModule:
             onnx_model=self.numpy_model.onnx_model,
         )
 
+        adapter = PowerOfTwoScalingRoundPBSAdapter(quantized_module)
+        # Apply the round PBS optimization if possible
+        adapter.process()
+
         self._process_input_quantizers(quantized_module, calibration_data)
 
         return quantized_module
diff --git a/src/concrete/ml/quantization/qat_quantizers.py b/src/concrete/ml/quantization/qat_quantizers.py
new file mode 100644
index 000000000..36a916546
--- /dev/null
+++ b/src/concrete/ml/quantization/qat_quantizers.py
@@ -0,0 +1,30 @@
+"""Custom Quantization Aware Training Brevitas quantizers."""
+from brevitas.quant.scaled_int import (
+    IntQuant,
+    MaxStatsScaling,
+    ParamFromRuntimePercentileScaling,
+    PerTensorPoTScaling8bit,
+    WeightQuantSolver,
+)
+from brevitas.quant.solver.act import ActQuantSolver
+
+# Note these classes are added here in order to isolate them from
+# the other modules, since the API doc generator has
+# an error when parsing them. Putting them in a separate
+# file allows us to ignore them during API doc generation
+
+
+# pylint: disable-next=too-many-ancestors
+class Int8ActPerTensorPoT(
+    IntQuant, ParamFromRuntimePercentileScaling, PerTensorPoTScaling8bit, ActQuantSolver
+):
+    """Quantization options for power-of-two scaling activations."""
+
+    _partialmethod = None
+
+
+# pylint: disable-next=too-many-ancestors
+class Int8WeightPerTensorPoT(IntQuant, MaxStatsScaling, PerTensorPoTScaling8bit, WeightQuantSolver):
+    """Quantization options for power-of-two scaling weights."""
+
+    _partialmethod = None
diff --git a/src/concrete/ml/quantization/quantized_module_passes.py b/src/concrete/ml/quantization/quantized_module_passes.py
new file mode 100644
index 000000000..517e2b15d
--- /dev/null
+++ b/src/concrete/ml/quantization/quantized_module_passes.py
@@ -0,0 +1,304 @@
+"""Optimization passes for QuantizedModules."""
+from collections import defaultdict
+from typing import DefaultDict, Dict, List, Optional, Tuple
+
+import numpy
+
+from ..common.debugging import assert_true
+from .base_quantized_op import QuantizedMixingOp, QuantizedOp
+from .quantized_module import QuantizedModule
+from .quantized_ops import (
+    QuantizedBrevitasQuant,
+    QuantizedConv,
+    QuantizedGemm,
+    QuantizedMatMul,
+    QuantizedRelu,
+)
+
+# A dictionary that contains for a quantized op a list of predecessor ops
+# Each predecessor op is stored along with its output tensor name
+PredecessorsType = DefaultDict[Optional[QuantizedOp], List[Tuple[Optional[QuantizedOp], str]]]
+
+# A list of optimizable patterns. For a "Mixing" op that supports rounding accumulators
+# we store a list of ops which contain information that allows us to
+# compute the integer scaling factor for the Mixing op.
+# The quantizer op of the input to the the Mixing op is stored in the second member of the tuple
+PatternDict = Dict[QuantizedMixingOp, Tuple[List[Optional[QuantizedOp]], Optional[QuantizedOp]]]
+
+
+class PowerOfTwoScalingRoundPBSAdapter:
+    """Detect neural network patterns that can be optimized with round PBS."""
+
+    SUPPORTED_ROUND_PBS_OPS = (QuantizedGemm, QuantizedMatMul, QuantizedConv)
+    SUPPORTED_ROUND_PBS_OP_PREDECESSOR = {
+        QuantizedBrevitasQuant: QuantizedRelu,
+        QuantizedRelu: QuantizedMixingOp,
+        QuantizedMixingOp: QuantizedBrevitasQuant,
+    }
+
+    def __init__(self, qmodule: QuantizedModule) -> None:
+        self._qmodule = qmodule
+        self._num_ignored_valid_patterns = 0
+
+    @property
+    def num_ignored_valid_patterns(self):
+        """Get the number of optimizable patterns that were ignored.
+
+        Patterns could be ignored since a number of rounding bits was
+        set manually through the compilation function.
+
+        Returns:
+            result (int): number of patterns that could be optimized but were not
+        """
+        return self._num_ignored_valid_patterns
+
+    def process(self) -> PatternDict:
+        """Analyze an ONNX graph and detect Gemm/Conv patterns that can use RoundPBS.
+
+        We want to detect a gemm/conv node whose weights/bias are Brevitas QAT, and whose
+        input is produced by a Brevitas QAT node that is applied on the output of
+        another Gemm/conv node. Optionally a Relu can be placed before this input
+        quantization node.
+
+        Nothing will be done if rounding is already specified.
+
+        Returns:
+            result (PatternDict): a dictionary containing for each Conv/Gemm node for which
+                round PBS can be applied based on power-of-two scaling factors
+        """
+
+        # The Pattern can be described as follows
+        # x = Quant(x) -> stored separately in the second member of the tuple in PatternDict
+        # .... the following ops are stored in the List of the PatternDict
+        # y = Gemm(x, w, b), with w, b produced by a Brevitas quant node
+        #     +---> This is the node for which roundPBS can be adjusted
+        # y = Relu(y)
+        # y = Quant(y)
+        # z = Gemm(y, w2, b2)  -> the output node of the pattern
+
+        self._num_ignored_valid_patterns = 0
+
+        predecessors = self.compute_op_predecessors()
+
+        valid_paths = self.detect_patterns(predecessors)
+
+        valid_paths = self.process_patterns(valid_paths)
+
+        return valid_paths
+
+    def compute_op_predecessors(self) -> PredecessorsType:
+        """Compute the predecessors for each QuantizedOp in a QuantizedModule.
+
+        Stores, for each quantized op, a list of quantized ops that produce its
+        inputs. Currently only the first input of the operations is considered
+        as it is, usually, the encrypted input.
+
+        Returns:
+            result (PredecessorsType): a dictionary containing a hierarchy of op
+                predecessors
+        """
+
+        # Initialize the list of predecessors with tensors that are graph inputs
+        predecessors: PredecessorsType = defaultdict(list)
+
+        for (node_inputs, node_op) in self._qmodule.quant_layers_dict.values():
+            # The first input node contains the encrypted data
+            enc_input_node = node_inputs[0]
+
+            assert_true(
+                enc_input_node in self._qmodule.quant_layers_dict
+                or enc_input_node in self._qmodule.ordered_module_input_names
+            )
+            pred = self._qmodule.quant_layers_dict.get(enc_input_node, (None, None))
+            # Get the quantized op that produces the current op's input
+            pred_with_output = (pred[1], enc_input_node)
+            predecessors[node_op].append(pred_with_output)
+        return predecessors
+
+    def match_path_pattern(
+        self,
+        predecessors: PredecessorsType,
+        nodes_in_path: List[Optional[QuantizedOp]],
+        input_producer_of_path: Optional[QuantizedOp],
+    ) -> bool:
+        """Determine if a pattern has the structure that makes it viable for roundPBS.
+
+        Args:
+            predecessors (PredecessorsType): Module predecessor operation list
+            nodes_in_path (List[QuantizedOp]): list of quantized ops in the pattern
+            input_producer_of_path (Optional[QuantizedOp]): operation that produces the input
+
+        Returns:
+            result (bool): whether the pattern can be optimized
+        """
+
+        # Test if the list of operations in this pattern has not the right length
+        if len(nodes_in_path) != 3:
+            return False
+
+        # If the input of this pattern is produced by a graph input then ignore it
+        # as graph inputs are not always quantized with QAT. QAT networks
+        # will have the input to the first gemm/conv op produced by a BrevitasQuant
+        # op and it will be valid pattern
+        if input_producer_of_path is None:
+            return False
+
+        for test_node in nodes_in_path:
+            # Check the operations in the pattern are chained properly
+            # for example if the Gemm op is preceded by a quantizer op, etc..
+            for pattern_first, pattern_second in self.SUPPORTED_ROUND_PBS_OP_PREDECESSOR.items():
+                pred_type = predecessors[test_node][0][0]
+                if isinstance(test_node, pattern_first) and not isinstance(
+                    pred_type, pattern_second
+                ):
+                    return False
+
+        return True
+
+    def detect_patterns(self, predecessors: PredecessorsType) -> PatternDict:
+        """Detect the patterns that can be optimized with roundPBS in the QuantizedModule.
+
+        Args:
+            predecessors (PredecessorsType): Module predecessor operation list
+
+        Returns:
+            result (PatternDict): list of optimizable patterns
+        """
+
+        valid_paths: PatternDict = {}
+
+        # pylint: disable-next=too-many-nested-blocks
+        for (_, node_op) in self._qmodule.quant_layers_dict.values():
+            # Only work with supported nodes that have a single
+            # encrypted input (not supporting enc x enc matmul)
+            if (
+                isinstance(node_op, self.SUPPORTED_ROUND_PBS_OPS)
+                and len(node_op.int_input_names) == 1
+            ):
+                prev_compatible_node_output = list(node_op.int_input_names)[0]
+                if len(predecessors[node_op]) == 1:
+                    back_node, back_node_output = predecessors[node_op][0]
+
+                    # A pattern is a sequence of Gemm/Conv -> Relu -> Quant
+                    # but we also need to store the Quant that quantizes
+                    # the Gemm/Conv's input
+                    nodes_in_path: List[Optional[QuantizedOp]] = []
+                    integer_node_input_quant: Optional[QuantizedOp] = None
+
+                    while back_node_output != prev_compatible_node_output:
+                        assert back_node is not None
+                        nodes_in_path.append(back_node)
+                        assert_true(
+                            back_node in predecessors,
+                            "Power of Two adapter: Error during graph traversal",
+                        )
+                        # If multiple ops produced this node, the pattern is not matched
+
+                        if len(predecessors[back_node]) == 1:
+                            back_node, back_node_output = predecessors[back_node][0]
+
+                            # Reached the previous integer node
+                            if back_node_output == prev_compatible_node_output:
+                                # The Gemm/Conv op that produces this integer node is the one
+                                # onto which we apply the roundPBS optimization
+                                nodes_in_path.append(back_node)
+                                list_pred_of_path = predecessors[back_node]
+                                if len(list_pred_of_path) == 1:
+                                    integer_node_input_quant = list_pred_of_path[0][0]
+
+                assert isinstance(node_op, QuantizedMixingOp)
+                if self.match_path_pattern(predecessors, nodes_in_path, integer_node_input_quant):
+                    # If rounding was manually set (usually globally for all layers)
+                    # the do not override the requested number of rounding bits
+                    # but keep statistics for testing purposes
+                    path_start_node = nodes_in_path[-1]
+                    assert isinstance(path_start_node, QuantizedMixingOp)
+                    if path_start_node.rounding_threshold_bits is not None:
+                        self._num_ignored_valid_patterns += 1
+                    else:
+                        valid_paths[path_start_node] = (nodes_in_path, integer_node_input_quant)
+        return valid_paths
+
+    def process_patterns(self, valid_paths: PatternDict) -> PatternDict:
+        """Configure the rounding bits of roundPBS for the optimizable operations.
+
+        Args:
+            valid_paths (PatternDict): list of optimizable patterns
+
+        Returns:
+            result (PatternDict): list of patterns actually optimized with roundPBS
+        """
+
+        def integer_log2(value: float) -> Tuple[int, bool]:
+            """Compute the log2 of the value and tests if its an integer.
+
+            Args:
+                value (float): the value for which to take the log2
+
+            Returns:
+                result: The integer log2 and a bool indicating whether
+                    the input value was an integer power of two
+            """
+            log2_value = int(numpy.rint(numpy.log2(value)))
+            # Check that the integer power of two is close to the original value
+            # with a small percentage tolerance
+            if numpy.isclose(numpy.power(2.0, log2_value), value, rtol=0.01):
+                return log2_value, True
+            return 0, False
+
+        invalid_paths: List[QuantizedMixingOp] = []
+        for path_start_node, (path, path_input_quant) in valid_paths.items():
+            # Placeholders
+            scale_input, scale_output, scale_weights = None, None, None
+            # Populate placeholders
+            for node in path:
+                if isinstance(node, self.SUPPORTED_ROUND_PBS_OPS):
+                    # Get the scale of the input of the Gemm/Conv node
+                    # and of its weights
+                    assert path_input_quant is not None
+                    scale_input = path_input_quant.constant_inputs[1]
+                    scale_weights = node.constant_inputs[1].quantizer.scale
+                elif isinstance(node, QuantizedBrevitasQuant):
+                    # Get the output scale that will be used to
+                    # compute the compounded scale factor of the
+                    # node that will apply roundPBS
+                    scale_output = node.constant_inputs[1]
+
+            # Check placeholders
+            assert scale_input is not None, (
+                "Power of two adapter: Can not determine input scale of pattern",
+            )
+            assert scale_weights is not None, (
+                "Power of two adapter: Can not determine weight scale of pattern",
+            )
+            assert scale_output is not None, (
+                "Power of two adapter: Can not determine output scale of pattern",
+            )
+
+            # Check if power of two
+            log2_input, ok_input = integer_log2(scale_input)
+            log2_weights, ok_weights = integer_log2(scale_weights)
+            log2_output, ok_output = integer_log2(scale_output)
+
+            # Modify rounding
+            if ok_input and ok_weights and ok_output:
+                assert_true(
+                    path_start_node.rounding_threshold_bits is None,
+                    "Power of two adapter: a global rounding configuration was unexpected here",
+                )
+                # The total scale factor is multiplied with the accumulator
+                # but we want to use a division with a power-of-two (shift right)
+                # operation to perform the scaling. Thus the
+                # number of lsbs to round is the negative of the sum of log2
+                # of the scale factors
+                lsbs_to_round = -(log2_input + log2_weights - log2_output)
+                if lsbs_to_round > 0:
+                    path_start_node.rounding_threshold_bits = lsbs_to_round
+                    path_start_node.lsbs_to_remove = lsbs_to_round
+            else:
+                invalid_paths.append(path_start_node)
+
+        for node in invalid_paths:
+            valid_paths.pop(node)
+
+        return valid_paths
diff --git a/src/concrete/ml/quantization/quantized_ops.py b/src/concrete/ml/quantization/quantized_ops.py
index 85d33293f..d7397c8bf 100644
--- a/src/concrete/ml/quantization/quantized_ops.py
+++ b/src/concrete/ml/quantization/quantized_ops.py
@@ -163,6 +163,7 @@ def __init__(
             f"{self._impl_for_op_named} if weights are provided as the 'b' constant input.",
         )
 
+    # pylint: disable-next=too-many-statements
     def q_impl(
         self,
         *q_inputs: ONNXOpInputOutputType,
@@ -183,7 +184,7 @@ def q_impl(
 
         q_input: QuantizedArray = prepared_inputs[0]
         q_weights: QuantizedArray = prepared_inputs[1]
-        q_bias: Optional[numpy.ndarray] = (
+        q_bias: Optional[QuantizedArray] = (
             None if len(prepared_inputs) == 2 or beta == 0 else prepared_inputs[2]
         )
 
@@ -263,13 +264,23 @@ def q_impl(
                 # Make mypy happy
                 assert q_bias is not None
                 # Reshape the biases to broadcast them to each neuron
-                out_zp = out_zp + q_bias / (-m_matmul)
+                bias_out = q_bias.values if isinstance(q_bias, QuantizedArray) else q_bias
+                out_zp = out_zp + bias_out / (-m_matmul)
 
             # We identify terms in the above equation to determine what
             # the scale/zero-point of the in-the-clear quantizer should be
             # to properly de-quantize numpy_q_out
             return self.make_output_quant_parameters(numpy_q_out, m_matmul, out_zp)
 
+        # Integer biases are only supported for Brevitas QAT which sets is_precomputed_qat to true
+        # These biases are produced by QuantizedBrevitasQuant ops
+        if q_bias is not None and q_bias.quantizer.is_precomputed_qat:
+            # Make sure the scale was correctly matching during training
+            # The bias scale should be the same scale as the one of the weights * inputs
+            assert q_bias.quantizer.scale is not None
+            assert numpy.isclose(q_bias.quantizer.scale, m_matmul)
+            numpy_q_out += q_bias.qvalues
+
         with tag(self.op_instance_name + ".matmul_rounding"):
             # Apply Concrete rounding (if relevant)
             numpy_q_out = self.cnp_round(numpy_q_out, calibrate_rounding)
@@ -280,9 +291,9 @@ def q_impl(
 
         numpy_q_out = m_matmul * numpy_q_out
 
-        if q_bias is not None:
+        if q_bias is not None and not q_bias.quantizer.is_precomputed_qat:
             # The bias is handled as a float and will be fused
-            numpy_q_out = numpy_q_out + q_bias
+            numpy_q_out = numpy_q_out + q_bias.values
 
         # Return the float values, so that Concrete can fuse any following float operations
         # We also keep track of the scaling factor and zero-point, since these will be
@@ -595,6 +606,7 @@ def __init__(
             " standard",
         )
 
+    # pylint: disable-next=too-many-statements
     def q_impl(
         self,
         *q_inputs: ONNXOpInputOutputType,
@@ -628,7 +640,7 @@ def q_impl(
         )
         q_input: QuantizedArray = prepared_inputs[0]
         q_weights: QuantizedArray = prepared_inputs[1]
-        q_bias: Optional[numpy.ndarray] = None if len(prepared_inputs) == 2 else prepared_inputs[2]
+        q_bias: Optional[QuantizedArray] = None if len(prepared_inputs) == 2 else prepared_inputs[2]
 
         in_channels = q_input.values.shape[1]
         weight_channels = q_weights.values.shape[1]
@@ -742,13 +754,20 @@ def q_impl(
             out_zp: Union[int, numpy.ndarray] = sum_weights - final_term
             if q_bias is not None:
                 # Reshape the biases to broadcast them to each channel
-                out_zp = out_zp - q_bias.reshape((1, -1, 1, 1)) / m_matmul
+                out_zp = out_zp - q_bias.values.reshape((1, -1, 1, 1)) / m_matmul
 
             # We identify terms in the above equation to determine what
             # the scale/zero-point of the in-the-clear quantizer should be
             # to properly de-quantize numpy_q_out
             return self.make_output_quant_parameters(numpy_q_out, m_matmul, out_zp)
 
+        if q_bias is not None and q_bias.quantizer.is_precomputed_qat:
+            # Make sure the scale was correctly matching during training
+            # The bias scale should be the same scale as the one of the weights * inputs
+            assert q_bias.quantizer.scale is not None
+            assert numpy.isclose(q_bias.quantizer.scale, m_matmul)
+            numpy_q_out += q_bias.qvalues.reshape((1, -1, 1, 1))
+
         with tag(self.op_instance_name + ".conv_rounding"):
             # Apply Concrete rounding (if relevant)
             numpy_q_out = self.cnp_round(numpy_q_out, calibrate_rounding)
@@ -759,10 +778,10 @@ def q_impl(
         # Rescale from scale=scale_inputs x scale_outputs to output scale
         numpy_q_out = m_matmul * numpy_q_out
 
-        if q_bias is not None:
+        if q_bias is not None and not q_bias.quantizer.is_precomputed_qat:
             # The bias addition is handled in float and will be fused into a TLU
             # Reshape the biases to broadcast them to each channel
-            numpy_q_out = numpy_q_out + q_bias.reshape((1, -1, 1, 1))  # bias_part
+            numpy_q_out = numpy_q_out + q_bias.values.reshape((1, -1, 1, 1))  # bias_part
 
         # And return as a QuantizedArray initialized from the float data, keeping
         # track of the quantization parameters
diff --git a/src/concrete/ml/quantization/quantizers.py b/src/concrete/ml/quantization/quantizers.py
index 73e9d145d..c3317cea4 100644
--- a/src/concrete/ml/quantization/quantizers.py
+++ b/src/concrete/ml/quantization/quantizers.py
@@ -7,6 +7,7 @@
 
 import numpy
 
+from ..common import utils
 from ..common.debugging import assert_true
 from ..common.serialization.dumpers import dump, dumps
 
@@ -745,7 +746,10 @@ def quant(self, values: numpy.ndarray) -> numpy.ndarray:
         assert self.offset is not None
         assert self.scale is not None
 
-        qvalues = numpy.rint(values / self.scale + self.zero_point)
+        if utils.QUANT_ROUND_LIKE_ROUND_PBS:
+            qvalues = numpy.floor(values / self.scale + self.zero_point + 0.5)
+        else:
+            qvalues = numpy.rint(values / self.scale + self.zero_point)
 
         # Clipping can be performed for PTQ and for precomputed (for now only Brevitas) QAT
         # (where quantizer parameters are available in ONNX layers).
diff --git a/src/concrete/ml/sklearn/qnn.py b/src/concrete/ml/sklearn/qnn.py
index b74dab5bb..7e4d94823 100644
--- a/src/concrete/ml/sklearn/qnn.py
+++ b/src/concrete/ml/sklearn/qnn.py
@@ -31,6 +31,7 @@
     "activation_function",
     "quant_narrow",
     "quant_signed",
+    "power_of_two_scaling",
 ]
 
 # skorch's special attribute prefixes, which can be found in:
diff --git a/src/concrete/ml/sklearn/qnn_module.py b/src/concrete/ml/sklearn/qnn_module.py
index 8710bcaaa..bd65831cc 100644
--- a/src/concrete/ml/sklearn/qnn_module.py
+++ b/src/concrete/ml/sklearn/qnn_module.py
@@ -5,10 +5,11 @@
 import numpy
 import torch
 import torch.nn.utils.prune as pruning
+from brevitas.quant.scaled_int import Int8ActPerTensorFloat, Int8WeightPerTensorFloat, IntBias
 from torch import nn
 
 from ..common.debugging import assert_true
-from ..common.utils import MAX_BITWIDTH_BACKWARD_COMPATIBLE
+from ..quantization.qat_quantizers import Int8ActPerTensorPoT, Int8WeightPerTensorPoT
 
 
 class SparseQuantNeuralNetwork(nn.Module):
@@ -27,13 +28,15 @@ def __init__(
         n_layers: int,
         n_outputs: int,
         n_hidden_neurons_multiplier: int = 4,
-        n_w_bits: int = 3,
-        n_a_bits: int = 3,
-        n_accum_bits: int = MAX_BITWIDTH_BACKWARD_COMPATIBLE,
+        n_w_bits: int = 4,
+        n_a_bits: int = 4,
+        # No pruning by default as roundPBS keeps the PBS precision low
+        n_accum_bits: int = 32,
         n_prune_neurons_percentage: float = 0.0,
         activation_function: Type = nn.ReLU,
         quant_narrow: bool = False,
         quant_signed: bool = True,
+        power_of_two_scaling: bool = True,  # Default to true: use roundPBS to speed up the NNs
     ):
         """Sparse Quantized Neural Network constructor.
 
@@ -60,6 +63,8 @@ def __init__(
                 (e.g a 2-bits signed quantization uses [-1, 0, 1] instead of [-2, -1, 0, 1]).
             quant_signed (bool): Whether this network should quantize the values using signed
                 integers.
+            power_of_two_scaling (bool): Force quantization scales to be a power of two
+                to enable inference speed optimizations. Defaults to True
 
         Raises:
             ValueError: If the parameters have invalid values or the computed accumulator bit-width
@@ -81,6 +86,8 @@ def __init__(
         if n_w_bits <= 0 or n_a_bits <= 0:
             raise ValueError("The weight & activation quantization bit-width cannot be less than 1")
 
+        high_input_bitwidth = False  # power_of_two_scaling and activation_function is nn.ReLU
+
         for idx in range(n_layers):
             out_features = (
                 n_outputs if idx == n_layers - 1 else int(input_dim * n_hidden_neurons_multiplier)
@@ -88,10 +95,11 @@ def __init__(
 
             quant_name = f"quant{idx}"
             quantizer = qnn.QuantIdentity(
-                bit_width=n_a_bits,
+                bit_width=8 if high_input_bitwidth else n_a_bits,
                 return_quant_tensor=True,
                 narrow_range=quant_narrow,
                 signed=quant_signed,
+                act_quant=Int8ActPerTensorPoT if power_of_two_scaling else Int8ActPerTensorFloat,
             )
 
             layer_name = f"fc{idx}"
@@ -100,10 +108,13 @@ def __init__(
                 out_features,
                 True,
                 weight_bit_width=n_w_bits,
-                bias_quant=None,
+                bias_quant=IntBias if power_of_two_scaling else None,
                 weight_narrow_range=quant_narrow,
                 narrow_range=quant_narrow,
                 signed=quant_signed,
+                weight_quant=Int8WeightPerTensorPoT
+                if power_of_two_scaling
+                else Int8WeightPerTensorFloat,
             )
 
             self.features.add_module(quant_name, quantizer)
diff --git a/tests/quantization/test_quantized_ops.py b/tests/quantization/test_quantized_ops.py
index 7431436ef..d3e7a15ee 100644
--- a/tests/quantization/test_quantized_ops.py
+++ b/tests/quantization/test_quantized_ops.py
@@ -468,13 +468,14 @@ def test_all_gemm_ops(
     # Quantize the inputs and weights
     q_inputs = QuantizedArray(n_bits, inputs)
     q_weights = QuantizedArray(n_bits, weights, is_signed=is_signed)
+    q_bias = QuantizedArray(n_bits, bias)
 
     # 1- Test our QuantizedGemm layer
     q_gemm = QuantizedGemm(
         n_bits,
         OP_DEBUG_NAME + "QuantizedGemm",
         int_input_names={"0"},
-        constant_inputs={"b": q_weights, "c": bias},
+        constant_inputs={"b": q_weights, "c": q_bias},
     )
     q_gemm.produces_graph_output = produces_output
 
@@ -532,7 +533,7 @@ def test_all_gemm_ops(
         n_bits,
         OP_DEBUG_NAME + "QuantizedGemm",
         int_input_names={"0"},
-        constant_inputs={"b": q_weights, "c": bias},
+        constant_inputs={"b": q_weights, "c": q_bias},
         alpha=1,
         beta=0,
     )
@@ -670,6 +671,7 @@ def test_identity_op(x, n_bits):
     ],
 )
 @pytest.mark.parametrize("produces_output", [True, False])
+# pylint: disable-next=too-many-locals
 def test_quantized_conv(params, n_bits, produces_output, check_r2_score, check_float_arrays_equal):
     """Test the quantized convolution operator."""
 
@@ -694,13 +696,14 @@ def test_quantized_conv(params, n_bits, produces_output, check_r2_score, check_f
     # Create quantized data
     q_input = QuantizedArray(n_bits, net_input, is_signed=False)
     q_weights = QuantizedArray(n_bits, weights, is_signed=True)
+    q_bias = QuantizedArray(n_bits, biases)
 
     # Create the operator, specifying weights & biases as constants
     q_op = QuantizedConv(
         n_bits,
         OP_DEBUG_NAME + "QuantizedConv",
         int_input_names={"0"},
-        constant_inputs={1: q_weights, 2: biases},
+        constant_inputs={1: q_weights, 2: q_bias},
         strides=strides,
         pads=pads,
         kernel_shape=(weights.shape[2], weights.shape[3]),
diff --git a/tests/sklearn/test_dump_onnx.py b/tests/sklearn/test_dump_onnx.py
index 2d51a6dab..00b22c4a9 100644
--- a/tests/sklearn/test_dump_onnx.py
+++ b/tests/sklearn/test_dump_onnx.py
@@ -2,6 +2,7 @@
 
 
 import warnings
+from functools import partial
 
 import numpy
 import onnx
@@ -11,6 +12,7 @@
 from concrete.ml.common.utils import is_model_class_in_a_list
 from concrete.ml.pytest.utils import get_model_name, sklearn_models_and_datasets
 from concrete.ml.sklearn import get_sklearn_tree_models
+from concrete.ml.sklearn.qnn import NeuralNetClassifier, NeuralNetRegressor
 
 # Remark that the dump tests for torch module is directly done in test_compile_torch.py
 
@@ -91,6 +93,29 @@ def test_dump(
     if parameters.get("n_classes", 2) != 2 and model_name in ["LinearSVC", "LogisticRegression"]:
         return
 
+    if model_name == "NeuralNetClassifier":
+        model_class = partial(
+            NeuralNetClassifier,
+            module__n_layers=3,
+            module__power_of_two_scaling=False,
+            max_epochs=1,
+            verbose=0,
+            callbacks="disable",
+        )
+    elif model_name == "NeuralNetRegressor":
+        model_class = partial(
+            NeuralNetRegressor,
+            module__n_layers=3,
+            module__n_w_bits=2,
+            module__n_a_bits=2,
+            module__n_accum_bits=7,  # Stay with 7 bits for test exec time
+            module__n_hidden_neurons_multiplier=1,
+            module__power_of_two_scaling=False,
+            max_epochs=1,
+            verbose=0,
+            callbacks="disable",
+        )
+
     n_classes = parameters.get("n_classes", 2)
 
     # Ignore long lines here
diff --git a/tests/sklearn/test_qnn.py b/tests/sklearn/test_qnn.py
index 1404dedb1..c0495db84 100644
--- a/tests/sklearn/test_qnn.py
+++ b/tests/sklearn/test_qnn.py
@@ -11,11 +11,14 @@
 from sklearn.preprocessing import StandardScaler
 from torch import nn
 
+from concrete.ml.common import utils
 from concrete.ml.common.utils import (
     MAX_BITWIDTH_BACKWARD_COMPATIBLE,
     is_classifier_or_partial_classifier,
     is_regressor_or_partial_regressor,
 )
+from concrete.ml.quantization.base_quantized_op import QuantizedMixingOp
+from concrete.ml.quantization.post_training import PowerOfTwoScalingRoundPBSAdapter
 from concrete.ml.sklearn import get_sklearn_neural_net_models
 from concrete.ml.sklearn.qnn import NeuralNetClassifier, NeuralNetRegressor
 from concrete.ml.sklearn.qnn_module import SparseQuantNeuralNetwork
@@ -186,6 +189,7 @@ def test_compile_and_calib(
         "module__n_a_bits": 2,
         "module__n_accum_bits": 5,
         "module__activation_function": activation_function,
+        "module__power_of_two_scaling": False,
         "max_epochs": 10,
         "verbose": 0,
     }
@@ -485,3 +489,139 @@ def test_serialization_unsupported_parameters(
 
     with pytest.raises(expected_error, match=expected_message):
         model.dumps()
+
+
+@pytest.mark.parametrize(
+    "activation_function",
+    [
+        pytest.param(nn.ReLU),
+        pytest.param(nn.Sigmoid),
+    ],
+)
+@pytest.mark.parametrize("num_layers", [2, 4])
+@pytest.mark.parametrize("model_class", [NeuralNetClassifier])
+@pytest.mark.parametrize("use_power_of_two_scaling", [True, False])
+def test_power_of_two_scaling(
+    activation_function,
+    model_class,
+    num_layers,
+    load_data,
+    use_power_of_two_scaling,
+    default_configuration,
+):
+    """Check that built-in neural networks can use roundPBS optimization."""
+
+    n_features = 10
+
+    # Get the data-set. The data generation is seeded in load_data.
+    x, y = load_data(
+        model_class,
+        n_samples=1000,
+        n_features=n_features,
+        n_redundant=0,
+        n_repeated=0,
+        n_informative=n_features,
+        n_classes=2,
+        class_sep=2,
+    )
+
+    # Perform a classic test-train split (deterministic by fixing the seed)
+    x_train, x_test, y_train, _ = train_test_split(
+        x,
+        y,
+        test_size=0.25,
+        random_state=numpy.random.randint(0, 2**15),
+    )
+
+    # Compute mean/stdev on training set and normalize both train and test sets with them
+    # Optimization algorithms for Neural networks work well on 0-centered inputs
+    normalizer = StandardScaler()
+    x_train = normalizer.fit_transform(x_train)
+    x_test = normalizer.transform(x_test)
+
+    # Configure a minimal neural network and train it quickly
+    params = {
+        "module__n_layers": num_layers,
+        "module__n_w_bits": 4,
+        "module__n_a_bits": 4,
+        "module__n_accum_bits": 32,
+        "module__activation_function": activation_function,
+        "module__power_of_two_scaling": use_power_of_two_scaling,
+        "max_epochs": 2,
+        "verbose": 0,
+    }
+
+    model = model_class(**params)
+
+    utils.QUANT_ROUND_LIKE_ROUND_PBS = True
+
+    # Train normally. This also converts the torch NN to a QuantizedModule
+    # and thus applies the PowerOfTwoScalingRoundPBSAdapter that
+    # detects and applies round PBS optimization
+    model.fit(x_train, y_train)
+
+    # Count the number of patterns that were optimized with roundPBS
+    num_round_pbs_layers = 0
+    for (_, node_op) in model.quantized_module_.quant_layers_dict.values():
+        if isinstance(node_op, QuantizedMixingOp):
+            num_round_pbs_layers += 1 if node_op.rounding_threshold_bits is not None else 0
+            assert node_op.rounding_threshold_bits == node_op.lsbs_to_remove
+
+    # Apply the PowerOfTwoScalingRoundPBSAdapter again. The second time
+    # the adapter will ignore already optimized patterns but report them
+    # as ignored.
+    adapter = PowerOfTwoScalingRoundPBSAdapter(model.quantized_module_)
+    round_pbs_patterns = adapter.process()
+
+    # The power-of-two optimization will only work
+    # when Relu activations are used and scaling factors are forced to be 2**s
+    if activation_function is nn.ReLU and use_power_of_two_scaling:
+        assert (
+            len(round_pbs_patterns) == 0
+        ), "Expected number of round PBS optimized patterns was not matched"
+        assert (
+            adapter.num_ignored_valid_patterns == num_layers - 1
+        ), "Expected number of ignored round PBS optimizable patterns was not matched"
+
+        y_pred_clear_round = model.predict(x_test, fhe="disable")
+
+        # Compile the model to ensure rounding is taken into account
+        # in compilation
+        model.compile(
+            x_train,
+            configuration=default_configuration,
+        )
+
+        # Compute the results with simulation, which uses the actual
+        # lookup tables.
+        y_pred_sim_round = model.predict(x_test, fhe="simulate")
+
+        # Ensure rounding was compiled in the circuit
+        # the number of rounding nodes should be equal
+        num_rounding_mlir = model.fhe_circuit.mlir.count(".round")
+
+        assert (
+            num_rounding_mlir == num_layers - 1
+        ), "Power-of-to adapter: Rounding nodes not found in MLIR"
+
+        # Remove rounding in the network to perform inference without the optimization.
+        # We expect a network that was optimized with the power-of-two adapter
+        # to be exactly correct to the non-optimized one
+        for (_, node_op) in model.quantized_module_.quant_layers_dict.values():
+            if isinstance(node_op, QuantizedMixingOp):
+                node_op.rounding_threshold_bits = None
+                node_op.lsbs_to_remove = None
+
+        # Predict with the unoptimized network
+        y_pred_clear_no_round = model.predict(x_test, fhe="disable")
+
+        # Compare the result with the optimized network with and without
+        # rounding. Tolerate at most 1 error
+        assert numpy.sum(y_pred_clear_round != y_pred_clear_no_round) <= 1
+        assert numpy.sum(y_pred_sim_round != y_pred_clear_no_round) <= 1
+    else:
+        # If the optimization is not expected to work, check that no patterns were
+        # detected
+        assert (
+            adapter.num_ignored_valid_patterns == 0
+        ), "Optimization performed but not expected for round PBS optimizable patterns"
diff --git a/tests/torch/test_brevitas_qat.py b/tests/torch/test_brevitas_qat.py
index ba60d9273..72e46d09a 100644
--- a/tests/torch/test_brevitas_qat.py
+++ b/tests/torch/test_brevitas_qat.py
@@ -1,48 +1,100 @@
 """Tests with brevitas quantization aware training."""
 
+from typing import Optional
+
 import brevitas.nn as qnn
 import numpy
 import pytest
 import torch
 import torch.utils
+from brevitas.quant import Int8ActPerTensorFloat, Int8WeightPerTensorFloat
+from brevitas.quant.scaled_int import IntBias
 from sklearn.datasets import load_digits
 from sklearn.model_selection import train_test_split
 from sklearn.preprocessing import StandardScaler
 from torch import nn
 from torch.utils.data import DataLoader, TensorDataset
 
+from concrete.ml.common import utils
 from concrete.ml.common.utils import (
     is_classifier_or_partial_classifier,
     is_regressor_or_partial_regressor,
 )
-from concrete.ml.pytest.torch_models import NetWithConstantsFoldedBeforeOps, TinyQATCNN
+from concrete.ml.pytest.torch_models import (
+    NetWithConstantsFoldedBeforeOps,
+    QuantCustomModel,
+    TinyQATCNN,
+)
+from concrete.ml.quantization.base_quantized_op import QuantizedMixingOp
+from concrete.ml.quantization.post_training import PowerOfTwoScalingRoundPBSAdapter
+from concrete.ml.quantization.qat_quantizers import Int8ActPerTensorPoT, Int8WeightPerTensorPoT
 from concrete.ml.sklearn import get_sklearn_neural_net_models
 from concrete.ml.sklearn.qnn_module import SparseQuantNeuralNetwork
 from concrete.ml.torch.compile import compile_brevitas_qat_model
 
 
-# This test is a known flaky
-# FIXME: https://github.com/zama-ai/concrete-ml-internal/issues/3933
-@pytest.mark.flaky
-@pytest.mark.parametrize("qat_bits", [3])
-@pytest.mark.parametrize("signed, narrow", [(True, False), (True, True), (False, False)])
-def test_brevitas_tinymnist_cnn(
-    qat_bits,
-    signed,
-    narrow,
-    default_configuration,
-    check_graph_input_has_no_tlu,
-    check_graph_output_has_no_tlu,
-    check_is_good_execution_for_cml_vs_circuit,
-):  # pylint: disable=too-many-statements, too-many-locals
-    """Train, execute and test a QAT CNN on a small version of MNIST."""
+def forward_test_torch(net, test_loader):
+    """Test the network: measure accuracy on the test set.
+
+    Args:
+        test_loader: the test loader
+
+    Returns:
+        res: the number of correctly classified test examples
+
+    """
+
+    # Freeze normalization layers
+    net.eval()
+
+    all_y_pred = numpy.zeros((len(test_loader)), dtype=numpy.int64)
+    all_targets = numpy.zeros((len(test_loader)), dtype=numpy.int64)
+
+    # Iterate over the batches
+    idx = 0
+    for data, target in test_loader:
+        # Accumulate the ground truth labels
+        endidx = idx + target.shape[0]
+        all_targets[idx:endidx] = target.numpy()
+
+        # Run forward and get the raw predictions first
+        raw_pred = net(data).detach().numpy()
+
+        # Get the predicted class id, handle NaNs
+        if numpy.any(numpy.isnan(raw_pred)):
+            output = -1  # pragma: no cover
+        else:
+            output = raw_pred.argmax(1)
+
+        all_y_pred[idx:endidx] = output
+
+        idx += target.shape[0]
+
+    # Print out the accuracy as a percentage
+    n_correct = numpy.sum(all_targets == all_y_pred)
+    return n_correct
 
+
+def train_brevitas_network_tinymnist(is_cnn, qat_bits, signed, narrow, pot_scaling):
+    """Train a QAT network on tiny mnist.
+
+    Args:
+        is_cnn (bool): whether to train a CNN or a FC network
+        qat_bits (int): quantization bits
+        signed (bool): use signed quantization
+        narrow (bool): use brevitas narrow range quantization
+        pot_scaling (int): use power of two scaling quantization
+
+    Returns:
+        result (Tuple): the network, the dataset and the test data loader
+    """
     # And some helpers for visualization.
     x_all, y_all = load_digits(return_X_y=True)
 
     # The sklearn Digits data-set, though it contains digit images, keeps these images in vectors
     # so we need to reshape them to 2D first. The images are 8x8 px in size and monochrome
-    x_all = numpy.expand_dims(x_all.reshape((-1, 8, 8)), 1)
+    if is_cnn:
+        x_all = numpy.expand_dims(x_all.reshape((-1, 8, 8)), 1)
 
     x_train, x_test, y_train, y_test = train_test_split(
         x_all, y_all, test_size=0.25, shuffle=True, random_state=numpy.random.randint(0, 2**15)
@@ -77,7 +129,19 @@ def train_one_epoch(net, optimizer, train_loader):
 
     while not trained_ok:
         # Create the tiny CNN module with 10 output classes
-        net = TinyQATCNN(10, qat_bits, 4 if qat_bits <= 3 else 20, signed, narrow)
+        if is_cnn:
+            net = TinyQATCNN(10, qat_bits, 4 if qat_bits <= 3 else 20, signed, narrow, pot_scaling)
+        else:
+            if pot_scaling:
+                act_quant = Int8ActPerTensorPoT
+                weight_quant = Int8WeightPerTensorPoT
+                bias_quant = IntBias
+            else:
+                act_quant = Int8ActPerTensorFloat
+                weight_quant = Int8WeightPerTensorFloat
+                bias_quant = None
+
+            net = QuantCustomModel(64, 10, 100, qat_bits, act_quant, weight_quant, bias_quant)
 
         # Train a single epoch to have a fast test, accuracy should still be the same for both
         # FHE simulation and torch
@@ -90,14 +154,38 @@ def train_one_epoch(net, optimizer, train_loader):
             train_one_epoch(net, optimizer, train_dataloader)
 
         # Finally, disable pruning (sets the pruned weights to 0)
-        net.toggle_pruning(False)
+        if hasattr(net, "toggle_pruning"):
+            net.toggle_pruning(False)
 
-        torch_correct = net.test_torch(test_dataloader)
+        torch_correct = forward_test_torch(net, test_dataloader)
 
         # If number of correct results was zero, training failed and there were NaNs in the weights
         # Retrain while training is bad
         trained_ok = torch_correct > 0
 
+    return net, x_all, test_dataloader
+
+
+# This test is a known flaky
+# FIXME: https://github.com/zama-ai/concrete-ml-internal/issues/3933
+@pytest.mark.flaky
+@pytest.mark.parametrize("qat_bits", [3])
+@pytest.mark.parametrize("signed, narrow", [(True, False), (True, True), (False, False)])
+def test_brevitas_tinymnist_cnn(
+    qat_bits,
+    signed,
+    narrow,
+    default_configuration,
+    check_graph_input_has_no_tlu,
+    check_graph_output_has_no_tlu,
+    check_is_good_execution_for_cml_vs_circuit,
+):  # pylint: disable=too-many-statements, too-many-locals
+    """Train, execute and test a QAT CNN on a small version of MNIST."""
+
+    net, x_all, test_dataloader = train_brevitas_network_tinymnist(
+        True, qat_bits, signed, narrow, False
+    )
+
     def test_with_concrete(quantized_module, test_loader, use_fhe_simulation):
         """Test a neural network that is quantized and compiled with Concrete ML."""
 
@@ -149,7 +237,7 @@ def test_with_concrete(quantized_module, test_loader, use_fhe_simulation):
     # FIXME: https://github.com/zama-ai/concrete-ml-internal/issues/2550
     # assert abs(fhe_simulation_correct - torch_correct) <= numpy.ceil(0.01 * len(y_test))
 
-    assert fhe_s_correct.shape == torch_correct.shape
+    assert fhe_s_correct >= 0
 
     check_graph_input_has_no_tlu(q_module_simulated.fhe_circuit.graph)
     check_graph_output_has_no_tlu(q_module_simulated.fhe_circuit.graph)
@@ -405,3 +493,107 @@ def test_brevitas_constant_folding(default_configuration):
             torch_inputset=data,
             configuration=default_configuration,
         )
+
+
+@pytest.mark.parametrize("manual_rounding", [None, 3])
+@pytest.mark.parametrize("power_of_two", [True, False])
+@pytest.mark.parametrize("n_bits", [4])
+@pytest.mark.parametrize("is_cnn", [True, False])
+def test_brevitas_power_of_two(
+    default_configuration,
+    manual_rounding: Optional[int],
+    power_of_two: bool,
+    n_bits: int,
+    is_cnn: bool,
+):
+    """Test a custom QAT network that uses power-of-two scaling.
+
+    Test whether a network using power-of-two scaling quantization is imported
+    correctly and roundPBS is used. Test that the Concrete ML does not override
+    the user's round PBS configuration.
+    """
+
+    net, x_all, _ = train_brevitas_network_tinymnist(is_cnn, n_bits, True, False, power_of_two)
+
+    utils.QUANT_ROUND_LIKE_ROUND_PBS = True
+
+    # If rounding threshold is set -> nothing happens
+    # If Quantizer is not setup -> nothing happens
+    quantized_module = compile_brevitas_qat_model(
+        net.to("cpu"),
+        torch_inputset=x_all,
+        configuration=default_configuration,
+        rounding_threshold_bits=manual_rounding,
+    )
+
+    pot_should_be_applied = not manual_rounding and power_of_two
+    # Count the number of patterns that were optimized with roundPBS
+    num_round_pbs_layers = 0
+    for (_, node_op) in quantized_module.quant_layers_dict.values():
+        if isinstance(node_op, QuantizedMixingOp):
+            num_round_pbs_layers += 1 if node_op.rounding_threshold_bits is not None else 0
+            if pot_should_be_applied:
+                assert node_op.rounding_threshold_bits == node_op.lsbs_to_remove
+            elif manual_rounding:
+                # If manual rounding was set, LSBs_to_remove must be equal
+                # to the accumulator size minus the requested rounding_threshold_bits
+                assert node_op.rounding_threshold_bits == manual_rounding
+                assert node_op.produces_graph_output or node_op.lsbs_to_remove is not None
+
+    # The power-of-two optimization will only work
+    # when Relu activations are used and scaling factors are forced to be 2**s
+    if not pot_should_be_applied:
+        return
+
+    # Apply the PowerOfTwoScalingRoundPBSAdapter again. The second time
+    # the adapter will ignore already optimized patterns but report them
+    # as ignored.
+    adapter = PowerOfTwoScalingRoundPBSAdapter(quantized_module)
+    round_pbs_patterns = adapter.process()
+
+    assert (
+        len(round_pbs_patterns) == 0
+    ), "Expected number of round PBS optimized patterns was not matched"
+    # 3 layers
+    assert (
+        adapter.num_ignored_valid_patterns == 3 - 1
+    ), "Expected number of ignored round PBS optimizable patterns was not matched"
+
+    x_test = x_all[numpy.random.choice(len(x_all), 100), ::]
+
+    x_test_q = quantized_module.quantize_input(x_test)
+
+    y_pred_clear_round = numpy.argmax(
+        quantized_module.quantized_forward(x_test_q, fhe="disable"), axis=1
+    )
+
+    # Compute the results with simulation, which uses the actual
+    # lookup tables.
+    y_pred_sim_round = numpy.argmax(
+        quantized_module.quantized_forward(x_test_q, fhe="simulate"), axis=1
+    )
+
+    # Ensure rounding was compiled in the circuit
+    # the number of rounding nodes should be equal
+    num_rounding_mlir = quantized_module.fhe_circuit.mlir.count(".round")
+
+    assert num_rounding_mlir == 2, "Power-of-to adapter: Rounding nodes not found in MLIR"
+
+    # Remove rounding in the network to perform inference without the optimization.
+    # We expect a network that was optimized with the power-of-two adapter
+    # to be exactly correct to the non-optimized one
+    for (_, node_op) in quantized_module.quant_layers_dict.values():
+        if isinstance(node_op, QuantizedMixingOp):
+            node_op.rounding_threshold_bits = None
+            node_op.lsbs_to_remove = None
+
+    # Predict with the unoptimized network
+    y_pred_clear_no_round = numpy.argmax(
+        quantized_module.quantized_forward(x_test_q, fhe="disable"), axis=1
+    )
+
+    # # Compare the result with the optimized network and without
+    # # they should be equal
+
+    assert numpy.sum(y_pred_sim_round != y_pred_clear_round) == 0
+    assert numpy.sum(y_pred_clear_round != y_pred_clear_no_round) == 0
diff --git a/tests/torch/test_compile_torch.py b/tests/torch/test_compile_torch.py
index 7db365c05..be5f50af6 100644
--- a/tests/torch/test_compile_torch.py
+++ b/tests/torch/test_compile_torch.py
@@ -1224,7 +1224,7 @@ def test_compilation_functions_check_model_types(default_configuration):
             configuration=default_configuration,
         )
 
-    torch_model_qat = TinyQATCNN(5, 4, 10, True, False)
+    torch_model_qat = TinyQATCNN(5, 4, 10, True, False, False)
     with pytest.raises(
         AssertionError, match=".*must be imported using compile_brevitas_qat_model.*"
     ):
diff --git a/use_case_examples/llm/utility_functions.py b/use_case_examples/llm/utility_functions.py
index 0d9ccadce..46d479f15 100644
--- a/use_case_examples/llm/utility_functions.py
+++ b/use_case_examples/llm/utility_functions.py
@@ -32,7 +32,7 @@ def max_fhe_relu(q_x, axis=-1, keepdims=True):
         if keepdims:
             shape = list(result.shape)
             shape.insert(axis, 1)
-            result = result.reshape(shape)
+            result = result.reshape(tuple(shape))
 
     return result