Added a complete example to export and evaluate an aihwkit-lightning model using the base aihwkit

docs/source/user_guide.rst

@@ -233,4 +233,6 @@ to 0 will not result in chunking. This is the default behavior.
 .. code-block:: python
 
     from aihwkit_lightning.nn.export import export_to_aihwkit
-    aihwkit_model = export_to_aihwkit(model=analog_model, max_output_size=-1)
+    aihwkit_model = export_to_aihwkit(model=analog_model, max_output_size=-1)
+
+A complete example is provided `here <https://github.com/IBM/aihwkit-lightning/tree/main/examples/aihwkit_evaluation>`_.

examples/aihwkit_evaluation/README.md

@@ -0,0 +1,4 @@
+# AIHWKIT Evaluation
+In this example, a model is converted to an analog-equivalent represenation using the aihwkit-lightning. Then, it is exported to an aihwkit representation which supports evaluation using a statistical inference model. Programming noise is applied using `drift_analog_weights()`.
+
+Note: To run this example, both aihwkit and aihwkit-lightning must be installed.

examples/aihwkit_evaluation/evaluation.py

@@ -0,0 +1,37 @@
+# -*- coding: utf-8 -*-
+
+# (C) Copyright 2024 IBM. All Rights Reserved.
+#
+# This code is licensed under the Apache License, Version 2.0. You may
+# obtain a copy of this license in the LICENSE.txt file in the root directory
+# of this source tree or at http://www.apache.org/licenses/LICENSE-2.0.
+#
+# Any modifications or derivative works of this code must retain this
+# copyright notice, and modified files need to carry a notice indicating
+# that they have been altered from the originals.
+
+"""Example on how to evaluate an AIHWKIT-Lightning model with AIHWKIT."""
+
+import torch
+from aihwkit.inference.noise.pcm import PCMLikeNoiseModel
+from aihwkit.inference.compensation.drift import GlobalDriftCompensation
+from model import resnet32
+from aihwkit_lightning.simulator.configs import TorchInferenceRPUConfig
+from aihwkit_lightning.nn.conversion import convert_to_analog
+from aihwkit_lightning.nn.export import export_to_aihwkit
+
+if __name__ == "__main__":
+    model = resnet32()
+    rpu_config = TorchInferenceRPUConfig()
+    model = convert_to_analog(model, rpu_config)
+    aihwkit_model = export_to_aihwkit(model=model, max_output_size=-1)
+    aihwkit_model.to(torch.float32)
+    for analog_tile in aihwkit_model.analog_tiles():
+        new_rpu_config = analog_tile.rpu_config
+        break
+
+    new_rpu_config.noise_model = PCMLikeNoiseModel(g_max=25.0)
+    new_rpu_config.drift_compensation = GlobalDriftCompensation()
+    aihwkit_model.replace_rpu_config(new_rpu_config)
+    aihwkit_model.eval()
+    aihwkit_model.drift_analog_weights(0.0)

examples/aihwkit_evaluation/model.py

@@ -0,0 +1,120 @@
+# -*- coding: utf-8 -*-
+
+# (C) Copyright 2024 IBM. All Rights Reserved.
+#
+# This code is licensed under the Apache License, Version 2.0. You may
+# obtain a copy of this license in the LICENSE.txt file in the root directory
+# of this source tree or at http://www.apache.org/licenses/LICENSE-2.0.
+#
+# Any modifications or derivative works of this code must retain this
+# copyright notice, and modified files need to carry a notice indicating
+# that they have been altered from the originals.
+
+"""Resnet32 model definition."""
+
+# pylint: skip-file
+
+import torch
+import torch.nn.functional as F
+from torch.nn import init
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+
+def _weights_init(m):
+    if isinstance(m, (torch.nn.Linear, torch.nn.Conv2d)):
+        init.kaiming_normal_(m.weight)
+
+
+class LambdaLayer(torch.nn.Module):
+    def __init__(self, lambd):
+        super(LambdaLayer, self).__init__()
+        self.lambd = lambd
+
+    def forward(self, x):
+        return self.lambd(x)
+
+
+class BasicBlock(torch.nn.Module):
+    expansion = 1
+
+    def __init__(self, in_planes, planes, stride=1, option="A"):
+        super(BasicBlock, self).__init__()
+        self.conv1 = torch.nn.Conv2d(
+            in_planes, planes, kernel_size=3, stride=stride, padding=1, bias=False
+        )
+        self.bn1 = torch.nn.BatchNorm2d(planes)
+        self.conv2 = torch.nn.Conv2d(
+            planes, planes, kernel_size=3, stride=1, padding=1, bias=False
+        )
+        self.bn2 = torch.nn.BatchNorm2d(planes)
+        self.shortcut = torch.nn.Sequential()
+        if stride != 1 or in_planes != planes:
+            if option == "A":
+                """
+                For CIFAR10 ResNet paper uses option A.
+                """
+                self.shortcut = LambdaLayer(
+                    lambda x: F.pad(
+                        x[:, :, ::2, ::2],
+                        (0, 0, 0, 0, planes // 4, planes // 4),
+                        "constant",
+                        0,
+                    )
+                )
+            elif option == "B":
+                self.shortcut = torch.nn.Sequential(
+                    torch.nn.Conv2d(
+                        in_planes,
+                        self.expansion * planes,
+                        kernel_size=1,
+                        stride=stride,
+                        bias=False,
+                    ),
+                    torch.nn.BatchNorm2d(self.expansion * planes),
+                )
+
+    def forward(self, x):
+        out = F.relu(self.bn1(self.conv1(x)))
+        out = self.bn2(self.conv2(out))
+        out += self.shortcut(x)
+        out = F.relu(out)
+        return out
+
+
+class ResNet(torch.nn.Module):
+    def __init__(self, block, num_blocks, n_classes=10):
+        super(ResNet, self).__init__()
+        self.in_planes = 16
+        self.conv1 = torch.nn.Conv2d(
+            3, 16, kernel_size=3, stride=1, padding=1, bias=False
+        )
+        self.bn1 = torch.nn.BatchNorm2d(16)
+        self.layer1 = self._make_layer(block, 16, num_blocks[0], stride=1)
+        self.layer2 = self._make_layer(block, 32, num_blocks[1], stride=2)
+        self.layer3 = self._make_layer(block, 64, num_blocks[2], stride=2)
+        self.linear = torch.nn.Linear(64, n_classes)
+        self.apply(_weights_init)
+
+    def _make_layer(self, block, planes, num_blocks, stride):
+        strides = [stride] + [1] * (num_blocks - 1)
+        layers = []
+        for stride in strides:
+            layers.append(block(self.in_planes, planes, stride))
+            self.in_planes = planes * block.expansion
+
+        return torch.nn.Sequential(*layers)
+
+    def forward(self, x):
+        out = F.relu(self.bn1(self.conv1(x)))
+        out = self.layer1(out)
+        out = self.layer2(out)
+        out = self.layer3(out)
+        out = F.avg_pool2d(out, out.size()[3])
+        out = out.view(out.size(0), -1)
+        out = self.linear(out)
+        return out
+
+
+def resnet32(n_classes=10):
+    return ResNet(BasicBlock, [5, 5, 5], n_classes=n_classes)