Merge pull request #1528 from AlbinSou/cil_exfree

CIL Examplar Free components

avalanche/models/__init__.py

@@ -27,3 +27,5 @@
 from .prompt import Prompt
 from .vit import create_model
 from .scr_model import *
+from .fecam import FeCAMClassifier
+from .cosine_layer import CosineIncrementalClassifier, CosineLinear

avalanche/models/cosine_layer.py

@@ -0,0 +1,180 @@
+#!/usr/bin/env python3
+import math
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import numpy as np
+
+from avalanche.models import DynamicModule
+
+
+"""
+Implementation of Cosine layer taken and modified from https://github.com/G-U-N/PyCIL
+"""
+
+
+class CosineLinear(nn.Module):
+    """
+    Cosine layer defined in
+    "Learning a Unified Classifier Incrementally via Rebalancing"
+    by Saihui Hou et al.
+
+    Implementation modified from https://github.com/G-U-N/PyCIL
+
+    This layer is aimed at countering the task-recency bias by removing the bias
+    in the classifier and normalizing the weight and the input feature before
+    computing the weight-feature product
+    """
+
+    def __init__(self, in_features, out_features, sigma=True):
+        """
+        :param in_features: number of input features
+        :param out_features: number of classes
+        :param sigma: learnable output scaling factor
+        """
+        super().__init__()
+        self.in_features = in_features
+        self.out_features = out_features
+        self.weight = nn.Parameter(torch.Tensor(self.out_features, in_features))
+        if sigma:
+            self.sigma = nn.Parameter(torch.Tensor(1))
+        else:
+            self.register_parameter("sigma", None)
+        self.reset_parameters()
+
+    def reset_parameters(self):
+        stdv = 1.0 / math.sqrt(self.weight.size(1))
+        self.weight.data.uniform_(-stdv, stdv)
+        if self.sigma is not None:
+            self.sigma.data.fill_(1)
+
+    def forward(self, input):
+        out = F.linear(
+            F.normalize(input, p=2, dim=1), F.normalize(self.weight, p=2, dim=1)
+        )
+        if self.sigma is not None:
+            out = self.sigma * out
+
+        return out
+
+
+class SplitCosineLinear(nn.Module):
+    """
+    This class keeps two Cosine Linear layers, without sigma scaling,
+    and handles the sigma parameter that is common for the two of them.
+    One CosineLinear is for the old classes and the other
+    one is for the new classes
+    """
+
+    def __init__(self, in_features, out_features1, out_features2, sigma=True):
+        super(SplitCosineLinear, self).__init__()
+        self.in_features = in_features
+        self.out_features = out_features1 + out_features2
+        self.fc1 = CosineLinear(in_features, out_features1, False)
+        self.fc2 = CosineLinear(in_features, out_features2, False)
+        if sigma:
+            self.sigma = nn.Parameter(torch.Tensor(1))
+            self.sigma.data.fill_(1)
+        else:
+            self.register_parameter("sigma", None)
+
+    def forward(self, x):
+        out1 = self.fc1(x)
+        out2 = self.fc2(x)
+
+        out = torch.cat((out1, out2), dim=1)
+
+        if self.sigma is not None:
+            out = self.sigma * out
+
+        return out
+
+
+class CosineIncrementalClassifier(DynamicModule):
+    """
+    Equivalent to IncrementalClassifier but using the cosine layer
+    described in "Learning a Unified Classifier Incrementally via Rebalancing"
+    by Saihui Hou et al.
+    """
+
+    def __init__(self, in_features, num_classes=0):
+        """
+        :param in_features: Number of input features
+        :param num_classes: Number of initial classes (default=0)
+                            If set to more than 0, the initial logits
+                            will be mapped to the corresponding sequence of
+                            classes starting from 0.
+        """
+        super().__init__()
+        self.class_order = []
+        self.classes = set()
+
+        if num_classes == 0:
+            self.fc = None
+        else:
+            self.fc = CosineLinear(in_features, num_classes, sigma=True)
+            for i in range(num_classes):
+                self.class_order.append(i)
+            self.classes = set(range(5))
+
+        self.feature_dim = in_features
+
+    def adaptation(self, experience):
+        num_classes = len(experience.classes_in_this_experience)
+
+        new_classes = set(experience.classes_in_this_experience) - set(self.classes)
+
+        if len(new_classes) == 0:
+            # Do not adapt
+            return
+
+        self.classes = self.classes.union(new_classes)
+
+        for c in list(new_classes):
+            self.class_order.append(c)
+
+        max_index = len(self.class_order)
+
+        if self.fc is None:
+            self.fc = CosineLinear(self.feature_dim, max_index, sigma=True)
+            return
+
+        fc = self._generate_fc(self.feature_dim, max_index)
+
+        if isinstance(self.fc, CosineLinear):
+            # First exp self.fc is CosineLinear
+            # while it is SplitCosineLinear for subsequent exps
+            fc.fc1.weight.data = self.fc.weight.data
+            fc.sigma.data = self.fc.sigma.data
+        elif isinstance(self.fc, SplitCosineLinear):
+            prev_out_features1 = self.fc.fc1.out_features
+            fc.fc1.weight.data[:prev_out_features1] = self.fc.fc1.weight.data
+            fc.fc1.weight.data[prev_out_features1:] = self.fc.fc2.weight.data
+            fc.sigma.data = self.fc.sigma.data
+
+        del self.fc
+        self.fc = fc
+
+    def forward(self, x):
+        unmapped_logits = self.fc(x)
+
+        # Mask by default unseen classes
+        mapped_logits = (
+            torch.ones(len(unmapped_logits), np.max(self.class_order) + 1) * -1000
+        )
+        mapped_logits.to(x.device)
+
+        # Now map to classes
+        mapped_logits[:, self.class_order] = unmapped_logits
+
+        return mapped_logits
+
+    def _generate_fc(self, in_dim, out_dim):
+        fc = SplitCosineLinear(
+            in_dim, self.fc.out_features, out_dim - self.fc.out_features
+        )
+        return fc
+
+
+__all__ = ["CosineLinear", "CosineIncrementalClassifier"]