Merge pull request #591 from vivekmig/v0.3.1

Captum v0.3.1

.circleci/config.yml

@@ -102,7 +102,9 @@ commands:
     steps:
       - run:
           name: "Simple PIP install"
-          command: python -m pip install -e .[dev]
+          command: |
+            python -m pip install --upgrade pip
+            python -m pip install -e .[dev]
 
   py_3_7_setup:
     description: "Set python version to 3.7 and install pip and pytest"

.gitignore

@@ -114,3 +114,4 @@ website/static/js/*
 !website/static/js/code_block_buttons.js
 website/static/_sphinx-sources/
 node_modules
+captum/insights/attr_vis/widget/static

README.md

@@ -7,7 +7,7 @@
 [![CircleCI](https://circleci.com/gh/pytorch/captum.svg?style=shield)](https://circleci.com/gh/pytorch/captum)
 
 Captum is a model interpretability and understanding library for PyTorch.
-Captum means comprehension in latin and contains general purpose implementations
+Captum means comprehension in Latin and contains general purpose implementations
 of integrated gradients, saliency maps, smoothgrad, vargrad and others for
 PyTorch models. It has quick integration for models built with domain-specific
 libraries such as torchvision, torchtext, and others.
@@ -175,12 +175,12 @@ Convergence Delta: tensor([2.3842e-07, -4.7684e-07])
 The algorithm outputs an attribution score for each input element and a
 convergence delta. The lower the absolute value of the convergence delta the better
 is the approximation. If we choose not to return delta,
-we can simply not provide `return_convergence_delta` input
+we can simply not provide the `return_convergence_delta` input
 argument. The absolute value of the returned deltas can be interpreted as an
 approximation error for each input sample.
 It can also serve as a proxy of how accurate the integral approximation for given
 inputs and baselines is.
-If the approximation error is large, we can try larger number of integral
+If the approximation error is large, we can try a larger number of integral
 approximation steps by setting `n_steps` to a larger value. Not all algorithms
 return approximation error. Those which do, though, compute it based on the
 completeness property of the algorithms.
@@ -224,7 +224,7 @@ in order to get per example average delta.
 
 
 Below is an example of how we can apply `DeepLift` and `DeepLiftShap` on the
-`ToyModel` described above. Current implementation of DeepLift supports only
+`ToyModel` described above. The current implementation of DeepLift supports only the
 `Rescale` rule.
 For more details on alternative implementations, please see the [DeepLift paper](https://arxiv.org/abs/1704.02685).
 
@@ -286,7 +286,7 @@ In order to smooth and improve the quality of the attributions we can run
 to smoothen the attributions by aggregating them for multiple noisy
 samples that were generated by adding gaussian noise.
 
-Here is an example how we can use `NoiseTunnel` with `IntegratedGradients`.
+Here is an example of how we can use `NoiseTunnel` with `IntegratedGradients`.
 
 ```python
 ig = IntegratedGradients(model)
@@ -338,7 +338,7 @@ It is an extension of path integrated gradients for hidden layers and holds the
 completeness property as well.
 
 It doesn't attribute the contribution scores to the input features
-but shows the importance of each neuron in selected layer.
+but shows the importance of each neuron in the selected layer.
 ```python
 lc = LayerConductance(model, model.lin1)
 attributions, delta = lc.attribute(input, baselines=baseline, target=0, return_convergence_delta=True)
@@ -412,6 +412,8 @@ See the [CONTRIBUTING](CONTRIBUTING.md) file for how to help out.
 ## Talks and Papers
 The slides of our presentation from NeurIPS 2019 can be found [here](docs/presentations/Captum_NeurIPS_2019_final.key)
 
+The slides of our presentation from KDD 2020 tutorial can be found [here](https://pytorch-tutorial-assets.s3.amazonaws.com/Captum_KDD_2020.pdf)
+
 ## References of Algorithms
 
 * `IntegratedGradients`, `LayerIntegratedGradients`: [Axiomatic Attribution for Deep Networks, Mukund Sundararajan et al. 2017](https://arxiv.org/abs/1703.01365) and [Did the Model Understand the Question?, Pramod K. Mudrakarta, et al. 2018](https://arxiv.org/abs/1805.05492)

captum/__init__.py

@@ -1,3 +1,3 @@
 #!/usr/bin/env python3
 
-__version__ = "0.3.0"
+__version__ = "0.3.1"

captum/_utils/common.py

@@ -249,6 +249,20 @@ def _expand_target(
     return target
 
 
+def _expand_feature_mask(
+    feature_mask: Union[Tensor, Tuple[Tensor, ...]], n_samples: int
+):
+    is_feature_mask_tuple = _is_tuple(feature_mask)
+    feature_mask = _format_tensor_into_tuples(feature_mask)
+    feature_mask_new = tuple(
+        feature_mask_elem.repeat_interleave(n_samples, dim=0)
+        if feature_mask_elem.size(0) > 1
+        else feature_mask_elem
+        for feature_mask_elem in feature_mask
+    )
+    return _format_output(is_feature_mask_tuple, feature_mask_new)
+
+
 def _expand_and_update_baselines(
     inputs: Tuple[Tensor, ...],
     n_samples: int,
@@ -317,6 +331,18 @@ def _expand_and_update_target(n_samples: int, kwargs: dict):
     kwargs["target"] = target
 
 
+def _expand_and_update_feature_mask(n_samples: int, kwargs: dict):
+    if "feature_mask" not in kwargs:
+        return
+
+    feature_mask = kwargs["feature_mask"]
+    if feature_mask is None:
+        return
+
+    feature_mask = _expand_feature_mask(feature_mask, n_samples)
+    kwargs["feature_mask"] = feature_mask
+
+
 @typing.overload
 def _format_output(
     is_inputs_tuple: Literal[True], output: Tuple[Tensor, ...]
@@ -354,6 +380,43 @@ def _format_output(
     return output if is_inputs_tuple else output[0]
 
 
+@typing.overload
+def _format_outputs(
+    is_multiple_inputs: Literal[False], outputs: List[Tuple[Tensor, ...]]
+) -> Union[Tensor, Tuple[Tensor, ...]]:
+    ...
+
+
+@typing.overload
+def _format_outputs(
+    is_multiple_inputs: Literal[True], outputs: List[Tuple[Tensor, ...]]
+) -> List[Union[Tensor, Tuple[Tensor, ...]]]:
+    ...
+
+
+@typing.overload
+def _format_outputs(
+    is_multiple_inputs: bool, outputs: List[Tuple[Tensor, ...]]
+) -> Union[Tensor, Tuple[Tensor, ...], List[Union[Tensor, Tuple[Tensor, ...]]]]:
+    ...
+
+
+def _format_outputs(
+    is_multiple_inputs: bool, outputs: List[Tuple[Tensor, ...]]
+) -> Union[Tensor, Tuple[Tensor, ...], List[Union[Tensor, Tuple[Tensor, ...]]]]:
+    assert isinstance(outputs, list), "Outputs must be a list"
+    assert is_multiple_inputs or len(outputs) == 1, (
+        "outputs should contain multiple inputs or have a single output"
+        f"however the number of outputs is: {len(outputs)}"
+    )
+
+    return (
+        [_format_output(len(output) > 1, output) for output in outputs]
+        if is_multiple_inputs
+        else _format_output(len(outputs[0]) > 1, outputs[0])
+    )
+
+
 def _run_forward(
     forward_func: Callable,
     inputs: Union[Tensor, Tuple[Tensor, ...]],

captum/_utils/models/linear_model/train.py

@@ -1,4 +1,5 @@
 import time
+import warnings
 from typing import Any, Callable, Dict, List, Optional
 
 import torch
@@ -286,10 +287,11 @@ def sklearn_train_linear_model(
     except ImportError:
         raise ValueError("sklearn is not available. Please install sklearn >= 0.23")
 
-    assert (
-        sklearn.__version__ >= "0.23.0"
-    ), "Must have sklearn version 0.23.0 or higher to use "
-    "sample_weight in Lasso regression."
+    if not sklearn.__version__ >= "0.23.0":
+        warnings.warn(
+            "Must have sklearn version 0.23.0 or higher to use "
+            "sample_weight in Lasso regression."
+        )
 
     num_batches = 0
     xs, ys, ws = [], [], []
@@ -323,7 +325,16 @@ def sklearn_train_linear_model(
     sklearn_model = reduce(
         lambda val, el: getattr(val, el), [sklearn] + sklearn_trainer.split(".")
     )(**construct_kwargs)
-    sklearn_model.fit(x, y, sample_weight=w, **fit_kwargs)
+    try:
+        sklearn_model.fit(x, y, sample_weight=w, **fit_kwargs)
+    except TypeError:
+        sklearn_model.fit(x, y, **fit_kwargs)
+        warnings.warn(
+            "Sample weight is not supported for the provided linear model!"
+            " Trained model without weighting inputs. For Lasso, please"
+            " upgrade sklearn to a version >= 0.23.0."
+        )
+
     t2 = time.time()
 
     # Convert weights to pytorch

captum/attr/_core/feature_ablation.py

@@ -346,6 +346,7 @@ def attribute(
                         eval_diff = (
                             initial_eval - modified_eval.reshape((-1, num_outputs))
                         ).reshape((-1, num_outputs) + (len(inputs[i].shape) - 1) * (1,))
+                        eval_diff = eval_diff.to(total_attrib[i].device)
                     if self.use_weights:
                         weights[i] += current_mask.float().sum(dim=0)
                     total_attrib[i] += (eval_diff * current_mask.to(attrib_type)).sum(

captum/attr/_core/gradient_shap.py

@@ -274,7 +274,7 @@ def attribute(
             nt,  # self
             inputs,
             nt_type="smoothgrad",
-            n_samples=n_samples,
+            nt_samples=n_samples,
             stdevs=stdevs,
             draw_baseline_from_distrib=True,
             baselines=baselines,

captum/attr/_core/kernel_shap.py

@@ -9,6 +9,7 @@
 from captum._utils.models.linear_model import SkLearnLinearRegression
 from captum._utils.typing import BaselineType, TargetType, TensorOrTupleOfTensorsGeneric
 from captum.attr._core.lime import Lime
+from captum.attr._utils.common import lime_n_perturb_samples_deprecation_decorator
 from captum.log import log_usage
 
 
@@ -72,14 +73,15 @@ def __init__(self, forward_func: Callable) -> None:
         )
 
     @log_usage()
+    @lime_n_perturb_samples_deprecation_decorator
     def attribute(  # type: ignore
         self,
         inputs: TensorOrTupleOfTensorsGeneric,
         baselines: BaselineType = None,
         target: TargetType = None,
         additional_forward_args: Any = None,
         feature_mask: Union[None, Tensor, Tuple[Tensor, ...]] = None,
-        n_perturb_samples: int = 25,
+        n_samples: int = 25,
         perturbations_per_eval: int = 1,
         return_input_shape: bool = True,
     ) -> TensorOrTupleOfTensorsGeneric:
@@ -213,9 +215,9 @@ def attribute(  # type: ignore
                         If None, then a feature mask is constructed which assigns
                         each scalar within a tensor as a separate feature.
                         Default: None
-            n_perturb_samples (int, optional):  The number of samples of the original
+            n_samples (int, optional):  The number of samples of the original
                         model used to train the surrogate interpretable model.
-                        Default: `50` if `n_perturb_samples` is not provided.
+                        Default: `50` if `n_samples` is not provided.
             perturbations_per_eval (int, optional): Allows multiple samples
                         to be processed simultaneously in one call to forward_fn.
                         Each forward pass will contain a maximum of
@@ -266,7 +268,7 @@ def attribute(  # type: ignore
             >>> ks = KernelShap(net)
             >>> # Computes attribution, with each of the 4 x 4 = 16
             >>> # features as a separate interpretable feature
-            >>> attr = ks.attribute(input, target=1, n_perturb_samples=200)
+            >>> attr = ks.attribute(input, target=1, n_samples=200)
 
             >>> # Alternatively, we can group each 2x2 square of the inputs
             >>> # as one 'interpretable' feature and perturb them together.
@@ -299,7 +301,7 @@ def attribute(  # type: ignore
             target=target,
             additional_forward_args=additional_forward_args,
             feature_mask=feature_mask,
-            n_perturb_samples=n_perturb_samples,
+            n_samples=n_samples,
             perturbations_per_eval=perturbations_per_eval,
             return_input_shape=return_input_shape,
         )

captum/attr/_core/layer/layer_gradient_shap.py

@@ -305,7 +305,7 @@ def attribute(
             nt,  # self
             inputs,
             nt_type="smoothgrad",
-            n_samples=n_samples,
+            nt_samples=n_samples,
             stdevs=stdevs,
             draw_baseline_from_distrib=True,
             baselines=baselines,