quadra 0.0.1__py3-none-any.whl → 2.2.7__py3-none-any.whl

quadra/utils/vit_explainability.py

@@ -0,0 +1,349 @@
+# Original Author: jacobgil
+# Title: Explainability for Vision Transformers
+# Source: https://github.com/jacobgil/vit-explain (MIT license)
+# Description: This is a heavily modified version of the original jacobgil code (the underlying math is still the same).
+from __future__ import annotations
+
+import math
+
+import numpy as np
+import torch
+from sklearn.linear_model._base import LinearClassifierMixin
+
+
+def rollout(
+    attentions: list[torch.Tensor], discard_ratio: float = 0.9, head_fusion: str = "mean", aspect_ratio: float = 1.0
+) -> np.ndarray:
+    """Apply rollout on Attention matrices.
+
+    Args:
+        attentions: List of Attention matrices coming from different blocks
+        discard_ratio: Percentage of elements to discard
+        head_fusion: Strategy of fusion of attention heads
+        aspect_ratio: Model inputs' width divided by height
+
+    Returns:
+        mask: Output mask, still needs a resize
+    """
+    result = torch.eye(attentions[0].size(-1))
+    with torch.no_grad():
+        for attention in attentions:
+            if head_fusion == "mean":
+                attention_heads_fused = attention.mean(dim=1)
+            elif head_fusion == "max":
+                attention_heads_fused = attention.max(dim=1)[0]
+            elif head_fusion == "min":
+                attention_heads_fused = attention.min(dim=1)[0]
+            else:
+                raise ValueError("Attention head fusion type Not supported")
+            # Drop the lowest attentions, but
+            # don't drop the class token
+            flat = attention_heads_fused.view(attention_heads_fused.size(0), -1)
+            _, indices = flat.topk(int(flat.size(-1) * discard_ratio), -1, False)
+            flat.scatter_(-1, indices, 0)
+            identity_matrix = torch.eye(attention_heads_fused.size(-1))
+            a = (attention_heads_fused + 1.0 * identity_matrix) / 2
+            a = a / a.sum(dim=-1).unsqueeze(1)
+            result = torch.matmul(a, result)
+    # Look at the total attention between the class token and the image patches
+    mask = result[:, 0, 1:]
+    batch_size = mask.size(0)
+    # TODO: Non squared input-size handling can be improved. Not easy though
+    height = math.floor((mask.size(-1) / aspect_ratio) ** 0.5)
+    total_size = mask.size(-1)
+    width = math.floor(total_size / height)
+    if mask.size(-1) > (height * width):
+        to_remove = mask.size(-1) - (height * width)
+        mask = mask[:, :-to_remove]
+    mask = mask.reshape(batch_size, height, width).numpy()
+    mask = mask / mask.max(axis=(1, 2), keepdims=True)
+
+    return mask
+
+
+class VitAttentionRollout:
+    """Attention gradient rollout class. Constructor registers hooks to the model's specified layers.
+    Only 4 layers by default given the high load on gpu. Best gradcams obtained using all blocks.
+
+    Args:
+        model: Model
+        attention_layer_names: On which layers to register the hook
+        head_fusion: Strategy of fusion for attention heads
+        discard_ratio: Percentage of elements to discard
+    """
+
+    def __init__(
+        self,
+        model: torch.nn.Module,
+        attention_layer_names: list[str] | None = None,
+        head_fusion: str = "mean",
+        discard_ratio: float = 0.9,
+    ):
+        if attention_layer_names is None:
+            attention_layer_names = [
+                "blocks.6.attn.attn_drop",
+                "blocks.7.attn.attn_drop",
+                "blocks.10.attn.attn_drop",
+                "blocks.11.attn.attn_drop",
+            ]
+        self.model = model
+        self.head_fusion = head_fusion
+        self.discard_ratio = discard_ratio
+        self.f_hook_handles: list[torch.utils.hooks.RemovableHandle] = []
+        for name, module in self.model.named_modules():
+            for layer_name in attention_layer_names:
+                if layer_name in name:
+                    self.f_hook_handles.append(module.register_forward_hook(self.get_attention))
+        self.attentions: list[torch.Tensor] = []
+
+    # pylint: disable=unused-argument
+    def get_attention(
+        self,
+        module: torch.nn.Module,
+        inpt: torch.Tensor,
+        out: torch.Tensor,
+    ) -> None:
+        """Hook to return attention.
+
+        Args:
+            module: Torch module
+            inpt: Input tensor
+            out: Output tensor, in this case the attention
+        """
+        self.attentions.append(out.detach().clone().cpu())
+
+    def __call__(self, input_tensor: torch.Tensor) -> np.ndarray:
+        """Called when the class instance is used as a function.
+
+        Args:
+            input_tensor: Input tensor
+
+        Returns:
+            out: Batch of output masks
+        """
+        self.attentions.clear()
+        with torch.no_grad():
+            self.model(input_tensor)
+        out = rollout(
+            self.attentions,
+            self.discard_ratio,
+            self.head_fusion,
+            aspect_ratio=(input_tensor.shape[-1] / input_tensor.shape[-2]),
+        )
+
+        return out
+
+
+############################################### GRAD ROLLOUT ##########################################################
+
+
+def grad_rollout(
+    attentions: list[torch.Tensor], gradients: list[torch.Tensor], discard_ratio: float = 0.9, aspect_ratio: float = 1.0
+) -> np.ndarray:
+    """Apply gradient rollout on Attention matrices.
+
+    Args:
+        attentions: Attention matrices
+        gradients: Target class gradient matrices
+        discard_ratio: Percentage of elements to discard
+        aspect_ratio: Model inputs' width divided by height
+
+    Returns:
+        mask: Output mask, still needs a resize
+    """
+    result = torch.eye(attentions[0].size(-1))
+    with torch.no_grad():
+        for attention, grad in zip(attentions, gradients):
+            weights = grad
+            attention_heads_fused = torch.mean((attention * weights), dim=1)
+            attention_heads_fused[attention_heads_fused < 0] = 0
+            # Drop the lowest attentions, but
+            # don't drop the class token
+            flat = attention_heads_fused.view(attention_heads_fused.size(0), -1)
+            _, indices = flat.topk(int(flat.size(-1) * discard_ratio), -1, False)
+            flat.scatter_(-1, indices, 0)
+            I = torch.eye(attention_heads_fused.size(-1))
+            a = (attention_heads_fused + 1.0 * I) / 2
+            a = a / a.sum(dim=-1).unsqueeze(1)
+            result = torch.matmul(a, result)
+    # Look at the total attention between the class token,
+    # and the image patches
+    mask = result[:, 0, 1:]
+    batch_size = mask.size(0)
+    # TODO: Non squared input-size handling can be improved. Not easy though
+    height = math.floor((mask.size(-1) / aspect_ratio) ** 0.5)
+    total_size = mask.size(-1)
+    width = math.floor(total_size / height)
+    if mask.size(-1) > (height * width):
+        to_remove = mask.size(-1) - (height * width)
+        mask = mask[:, :-to_remove]
+    mask = mask.reshape(batch_size, height, width).numpy()
+    mask = mask / mask.max(axis=(1, 2), keepdims=True)
+
+    return mask
+
+
+class VitAttentionGradRollout:
+    """Attention gradient rollout class. Constructor registers hooks to the model's specified layers.
+    Only 4 layers by default given the high load on gpu. Best gradcams obtained using all blocks.
+
+    Args:
+        model: Pytorch model
+        attention_layer_names: On which layers to register the hooks
+        discard_ratio: Percentage of elements to discard
+        classifier: Scikit-learn classifier. Leave it to None if model already has a classifier on top.
+    """
+
+    def __init__(  # pylint: disable=W0102
+        self,
+        model: torch.nn.Module,
+        attention_layer_names: list[str] | None = None,
+        discard_ratio: float = 0.9,
+        classifier: LinearClassifierMixin | None = None,
+        example_input: torch.Tensor | None = None,
+    ):
+        if attention_layer_names is None:
+            attention_layer_names = [
+                "blocks.6.attn.attn_drop",
+                "blocks.7.attn.attn_drop",
+                "blocks.10.attn.attn_drop",
+                "blocks.11.attn.attn_drop",
+            ]
+
+        if classifier is not None:
+            if example_input is None:
+                raise ValueError(
+                    "Must provide an input example to LinearModelPytorchWrapper when classifier is not None"
+                )
+            self.model = LinearModelPytorchWrapper(
+                backbone=model,
+                linear_classifier=classifier,
+                example_input=example_input,
+                device=next(model.parameters()).device,
+            )
+        else:
+            self.model = model  # type: ignore[assignment]
+
+        self.discard_ratio = discard_ratio
+        self.f_hook_handles: list[torch.utils.hooks.RemovableHandle] = []
+        self.b_hook_handles: list[torch.utils.hooks.RemovableHandle] = []
+        for name, module in self.model.named_modules():
+            for layer_name in attention_layer_names:
+                if layer_name in name:
+                    self.f_hook_handles.append(module.register_forward_hook(self.get_attention))
+                    self.b_hook_handles.append(module.register_backward_hook(self.get_attention_gradient))
+        self.attentions: list[torch.Tensor] = []
+        self.attention_gradients: list[torch.Tensor] = []
+        # Activate gradients
+        blocks_list = [x.split("blocks")[1].split(".attn")[0] for x in attention_layer_names]
+        for name, module in model.named_modules():
+            for p in module.parameters():
+                if "blocks" in name and any(x in name for x in blocks_list):
+                    p.requires_grad = True
+
+    # pylint: disable=unused-argument
+    def get_attention(
+        self,
+        module: torch.nn.Module,
+        inpt: torch.Tensor,
+        out: torch.Tensor,
+    ) -> None:
+        """Hook to return attention.
+
+        Args:
+            module: Torch module
+            inpt: Input tensor
+            out: Output tensor, in this case the attention
+        """
+        self.attentions.append(out.detach().clone().cpu())
+
+    # pylint: disable=unused-argument
+    def get_attention_gradient(
+        self,
+        module: torch.nn.Module,
+        grad_input: torch.Tensor,
+        grad_output: torch.Tensor,
+    ) -> None:
+        """Hook to return attention.
+
+        Args:
+            module: Torch module
+            grad_input: Gradients' input tensor
+            grad_output: Gradients' output tensor, in this case the attention
+        """
+        self.attention_gradients.append(grad_input[0].detach().clone().cpu())
+
+    def __call__(self, input_tensor: torch.Tensor, targets_list: list[int]) -> np.ndarray:
+        """Called when the class instance is used as a function.
+
+        Args:
+            input_tensor: Model's input tensor
+            targets_list: List of targets. If None, argmax is used
+
+        Returns:
+            out: Batch of output masks
+        """
+        self.attentions.clear()
+        self.attention_gradients.clear()
+
+        self.model.zero_grad(set_to_none=True)
+        self.model.to(input_tensor.device)
+        output = self.model(input_tensor).cpu()
+
+        class_mask = torch.zeros(output.shape)
+        if targets_list is None:
+            targets_list = output.argmax(dim=1)
+        class_mask[torch.arange(output.shape[0]), targets_list] = 1
+        loss = (output * class_mask).sum()
+        loss.backward()
+        out = grad_rollout(
+            self.attentions,
+            self.attention_gradients,
+            self.discard_ratio,
+            aspect_ratio=(input_tensor.shape[-1] / input_tensor.shape[-2]),
+        )
+
+        return out
+
+
+class LinearModelPytorchWrapper(torch.nn.Module):
+    """Pytorch wrapper for scikit-learn linear models.
+
+    Args:
+        backbone: Backbone
+        linear_classifier: ScikitLearn linear classifier model
+        example_input: Input example needed to obtain output shape
+        device: The device to use. Defaults to "cpu"
+    """
+
+    def __init__(
+        self,
+        backbone: torch.nn.Module,
+        linear_classifier: LinearClassifierMixin,
+        example_input: torch.Tensor,
+        device: torch.device,
+    ):
+        super().__init__()
+        self.device = device
+        self.backbone = backbone.to(device)
+        if not isinstance(linear_classifier, LinearClassifierMixin):
+            raise TypeError("Classifier is not of type LinearClassifierMixin.")
+        self.num_classes = len(linear_classifier.classes_)
+        self.linear_classifier = linear_classifier
+        with torch.no_grad():
+            output = self.backbone(example_input.to(device))
+            num_filters = output.shape[-1]
+
+        self.classifier = torch.nn.Linear(num_filters, self.num_classes).to(device)
+        self.classifier.weight.data = torch.from_numpy(linear_classifier.coef_).float()
+        self.classifier.bias.data = torch.from_numpy(linear_classifier.intercept_).float()
+
+    def forward(self, x):
+        if self.num_classes == 2:
+            class_one_probabilities = torch.nn.Sigmoid()(self.classifier(self.backbone(x)))
+            class_zero_probabilities = 1 - class_one_probabilities
+            two_class_probs = torch.cat((class_zero_probabilities, class_one_probabilities), dim=1)
+
+            return two_class_probs
+
+        return torch.nn.Softmax(dim=1)(self.classifier(self.backbone(x)))

quadra-2.2.7.dist-info/LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [2023] Orobix Srl
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.