Code for the TCAV ML interpretability project

Last update: Dec 27, 2022

Overview

Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)

Been Kim, Martin Wattenberg, Justin Gilmer, Carrie Cai, James Wexler, Fernanda Viegas, Rory Sayres

ICML Paper: https://arxiv.org/abs/1711.11279

What is TCAV?

Testing with Concept Activation Vectors (TCAV) is a new interpretability method to understand what signals your neural networks models uses for prediction.

What's special about TCAV compared to other methods?

Typical interpretability methods show importance weights in each input feature (e.g, pixel). TCAV instead shows importance of high level concepts (e.g., color, gender, race) for a prediction class - this is how humans communicate!

Typical interpretability methods require you to have one particular image that you are interested in understanding. TCAV gives an explanation that is generally true for a class of interest, beyond one image (global explanation).

For example, for a given class, we can show how much race or gender was important for classifications in InceptionV3. Even though neither race nor gender labels were part of the training input!

Cool, where do these concepts come from?

TCAV learns concepts from examples. For instance, TCAV needs a couple of examples of female, and something not female to learn a "gender" concept. We have tested a variety of concepts: color, gender, race, textures and many others.

Why use high level concepts instead of input features?

Humans think and communicate using concepts, and not using numbers (e.g., weights to each feature). When there are lots of numbers to combine and reason about (many features), it becomes harder and harder for humans to make sense of the information they are accounting for. TCAV instead delivers explanations in the way humans communicate to each other.

The consumer of the explanation may not know machine learning too well. Can they understand the explanation?

Yes. TCAV is designed to make sense to everyone - as long as they can understand the high level concept!

Sounds good. Do I need to change my network to use TCAV?

No. You don't need to change or retrain your network to use TCAV.

Installation

Tensorflow must be installed to use TCAV. But it isn't included in the TCAV pip package install_requires as a user may wish to use it with either the tensorflow or tensorflow-gpu package. So please pip install tensorflow or tensorflow-gpu as well as the tcav package.

pip install tcav

Requirements

See requirements.txt for a list of python dependencies used in testing TCAV. These will all be installed during pip installation of tcav with the exception of tensorflow, as mentioned above.

How to use TCAV

See Run TCAV.ipynb for step by step guide, after pip installing the tcav package.

mytcav = tcav.TCAV(sess,
                   target,
                   concepts,
                   bottlenecks,
                   act_gen,
                   alphas,
                   cav_dir=cav_dir,
                   num_random_exp=2)

results = mytcav.run()

TCAV for discrete models

We provide a simple example of how to run TCAV on models trained on discrete, non-image data. Please see

cd tcav/tcav_examples/discrete/

You can also find a Jupyter notebook for a model trained on KDD99 in here:

tcav/tcav_examples/discrete/kdd99_discrete_example.ipynb.

Requirements

tensorflow
numpy
Pillow
matplotlib
scikit-learn
scipy

How to run unit tests

python -m tcav.cav_test

python -m tcav.model_test

python -m tcav.tcav_test

python -m tcav.utils_test

How to create a new version of the pip package

Ensure the version in setup.py has been updated to a new version.
Run python setup.py bdist_wheel --python-tag py3 and python setup.py bdist_wheel --python-tag py2.
Run twine upload dist/* to upload the py2 and py3 pip packages to PyPi.

Comments

Random Images for TCAV for Diabetic Retinopathy application

Hi,

I had a question regarding application of TCAV for Diabetic Retinopathy (DR). My question is, for application of TCAV on DR, what do you choose as the random images. Are they completely random images as shown in paper or are the random images related to DR or healthy eye?

Also, how many random folders/experiments do you use for DR application? For the statistical significance testing?

opened by soumbane 9
Unable to run the tests, program and notebook
Hey there,

I was trying to use the code from the repository and i found that things were not working for me. Here are the things that are of concern to me :

Run TCAV.ipynb returns me an error saying No module named tcav_results despite installing tcav.

pip install -r requirements.txt throws an error because you used a single = sign instead of ==.

python -m tcav.cav_test throws me an error saying ModuleNotFoundError: No module named 'tcav_results'

Useful information :

python --version

Python 3.6.5

pip --version

pip 19.3.1

uname -a

Linux x1 4.15.0-70-generic #79~16.04.1-Ubuntu SMP Tue Nov 12 14:01:10 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux
opened by akhilpandey95 9
Which implementation of InceptionV3 should I be using?

Hi,

Which implementation of inceptionv3 should I be using to replicate your results? The 3 I have used did not have the names of the layers model.py is looking for. I'd really appreciate if you can point us to the code for inception.

Thanks!

opened by rahul003 9
hypothesis testing module

I could run tcav on zebra class with concept images. however couldn't find the hypothesis testing (Z-test/T-test) section of code in the repo. "If we can reject the null hypothesis of a TCAV score of 0.5, we can consider the resulting concept as related to the class prediction in a significant way. " does that mean consider only those concepts that get tcav score > 0.5 and also pass the hypothesis test?

opened by KavyaBPadmesh 6
question regarding model wrappers (xception model)

Hi I have a imagenet-pretrained xception network trained on histology images. I would like to be able to use tcav with this model; however, I'm unsure of how to determine the bottleneck layers that tcav asks for. 1) am i able to use this model for tcav and 2) do you have any suggestions on how to get the bottleneck tensors? I can get all the tensors for the network with graph.get_operations but I'm not sure if any or which of those are relevant.

Thank you for your help!

opened by justcho5 5
Question regarding directional derivative computation

Hi, While working with tcav code after reading the related paper " Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)" (https://arxiv.org/pdf/1711.11279.pdf), I had trouble understanding the computation of the directional derivative.

From my understanding of the paper, 'Conceptual sensitivity' is measured by the directional derivative of logit values with respect to activations at the layer of interest. Thus for scoring you take into account the proportions of image with positive directional derivative.

On the other hand in the code, as shown in the snippet below, sensitivity score seems to be the directional derivative of loss values with respect to activations at the layer of interest. https://github.com/tensorflow/tcav/blob/c7f55108e26c66fef9d21754d4d697ed6337ddc3/tcav/model.py#L54-L55

As the loss is a function of the logits, the gradient we obtain in the code depends of the loss function we choose. In your case with cross entropy (snippet below) using the article's notation we have :

This leads to a change in magnitude ( which is not used is the score) but also in terms of sign. https://github.com/tensorflow/tcav/blob/c7f55108e26c66fef9d21754d4d697ed6337ddc3/tcav/model.py#L173-L178

Though it seems that in the code you compensate the sign change by counting the number of negative gradients instead (last snippet). https://github.com/tensorflow/tcav/blob/c7f55108e26c66fef9d21754d4d697ed6337ddc3/tcav/tcav.py#L42-L60

Have I misunderstood either the paper or the code ? Or is there a specific reason to use loss instead of logit, as you still obtain the same score by changing sign ? Thank you for your answer !

opened by hugoych 5
Changes to random concept calculation and bug fixes
Hello! As discussed previously, here is a pull request for some changes I've been working on.

Change list

Files are now opened in binary mode for python3 compatibility

Raise ValueError on trying to run TCAV on one image. Things seem to break when this happens.

Remove transparency channel from images

Print results outputs summaries for multiple concepts, and optionally outputs a plot. See example below

Changed the way random scores are calculated. Instead of nominating a single random_counterpart to use as the positive set, all negative sets of images are used in turn as the positve set. This has a great effect on the mean of the random concept, however is very computationally expensive.

Removed random_counterpart arg for TCAV

Update tests for these changes

I understand you may not want random_concept removed, I can re-add it. What effect should it have? Also, maybe the change to random concepts should be optional due to the extra computation time required.

Another thing to look into here is to save computing time by noting that the TCAV score for random500_1 vs random500_0 should be the same as 1 - (random500_0 vs random500_1). I am currently unsure how to implement this however.

Sample output from print_results can be seen below
opened by Gareth001 5
TCAV with TensorFlow Object Detection API models

Hi, I run the repo without issues, and I was able to apply it to Keras models (image classification) successfully. Now I'm looking at how to apply it to Object Detection. But I'm pretty lost because the Object Detection API model is an SSDMetaArch file. If you use the *.pb file, don't provide access to backbone layers directly, and these layers are on Keras. Hence I can apply the Keras wrapper neither use one that follows a graph such as the examples provided in this repository. Any idea or suggestion on where I can start? Quickly add what I'm able to do:

Get the Keras classification backbone of the Object detector model Get the bottleneck of this model Get the input tensor of this model I cannot pull the loss function from the model, but I can write one. What not being able to do;

Get the prediction layer not sure if I need to get the prediction layer by individual bounding boxes I really appreciate any help you can provide. P.S. if more details or info is needed from my side. please let me know

opened by emmanuelol 4
ImageNet BaseURL updated.

The links requested to ImageNet returned a 404 error. fix: The BaseURL doesn't contain "text" anymore. Also, the urls had extra "/r/n" at the end, which was truncated.

opened by MatinTavakoli 4

Kernel crashing with an unknow reason

I am trying to use tcav with my CNN keras model, but my kernel keeps crashing for an unknown reason when launching tcav.run(). After checking the code seems that the error is presenting when calling compute_tcav_score in this line of tcav.py

i_up = self.compute_tcav_score( mymodel, target_class_for_compute_tcav_score, cav_concept, cav_instance, acts[target_class][cav_instance.bottleneck], activation_generator.get_examples_for_concept(target_class), run_parallel=run_parallel)

The wrapper I have created for my keras model is the following:

class KerasModelWrapper(model.ImageModelWrapper): """ ModelWrapper for keras models

    By default, assumes that your model contains one input node, one output head
and one loss function.
Computes gradients of the output layer in respect to a CAV.

Args:
    sess: Tensorflow session we will use for TCAV.
    model_path: Path to your model.h5 file, containing a saved trained
      model.
    labels_path: Path to a file containing the labels for your problem. It
      requires a .txt file, where every line contains a label for your
      model. You want to make sure that the order of labels in this file
      matches with the logits layers for your model, such that file[i] ==
      model_logits[i]
"""

def __init__(
    self,
    sess,
    model_path,
    labels_path,
    image_shape
):
    self.sess = sess
    self.model_name = 'cifar_CNN'
    super(KerasModelWrapper, self).__init__(image_shape)
    self.import_keras_model(model_path)
    self.labels = tf.io.gfile.GFile(labels_path).read().splitlines()

    # Construct gradient ops. Defaults to using the model's output layer
    self.y_input = tf.compat.v1.placeholder(tf.int64, shape=[None])
    self.loss = self.model.loss_functions[0](self.y_input,
                                             self.model.outputs[0])
    self._make_gradient_tensors()

def id_to_label(self, idx):
    return self.labels[idx]

def label_to_id(self, label):
    return self.labels.index(label)

def import_keras_model(self, saved_path):
    """Loads keras model, fetching bottlenecks, inputs and outputs."""
    self.ends = {}
    self.model = tf.keras.models.load_model(saved_path)
    self.get_bottleneck_tensors()
    self.get_inputs_and_outputs_and_ends()

def get_bottleneck_tensors(self):
    self.bottlenecks_tensors = {}
    layers = self.model.layers
    for layer in layers:
        if 'input' not in layer.name:
            self.bottlenecks_tensors[layer.name] = layer.output

def get_inputs_and_outputs_and_ends(self):
    self.ends['input'] = self.model.inputs[0]
    self.ends['prediction'] = self.model.outputs[0]

I don't know if can help but from the log I can retrieve only this information before the crash F ./tensorflow/core/util/bcast.h:111] Check failed: vec.size() == NDIMS (1 vs. 2)

What am I doing wrong?

opened by francescobodria 4

google.protobuf.message.DecodeError: Error parsing message

TCAV library -- there's a problem when importing in utils.py. Particularly the line when importing the results_pb2 module. I fixed it by changing "from tcav_results.results_pb2..." to "...tcav.tcav_results_pb2...".

Also there is an incompatibility with tf 2.0. I've tried by replacing tf. with tf.compat.v1. However, I then ran into a decode error, which i opened another issue on.

In addition, i get this error: File "./tcav/tcav/model.py", line 318, in import_graph graph_def = tf.compat.v1.GraphDef.FromString(tf.compat.v1.gfile.Open(saved_path, 'rb').read()) google.protobuf.message.DecodeError: Error parsing message

Do you know how I could resolve this problem? Thank you

opened by justcho5 4
Bump pillow from 8.2.0 to 9.3.0
Bumps pillow from 8.2.0 to 9.3.0.

Release notes

Sourced from pillow's releases.

9.3.0

https://pillow.readthedocs.io/en/stable/releasenotes/9.3.0.html

Changes

Initialize libtiff buffer when saving #6699 [@radarhere]

Limit SAMPLESPERPIXEL to avoid runtime DOS #6700 [@wiredfool]

Inline fname2char to fix memory leak #6329 [@nulano]

Fix memory leaks related to text features #6330 [@nulano]

Use double quotes for version check on old CPython on Windows #6695 [@hugovk]

GHA: replace deprecated set-output command with GITHUB_OUTPUT file #6697 [@nulano]

Remove backup implementation of Round for Windows platforms #6693 [@cgohlke]

Upload fribidi.dll to GitHub Actions #6532 [@nulano]

Fixed set_variation_by_name offset #6445 [@radarhere]

Windows build improvements #6562 [@nulano]

Fix malloc in _imagingft.c:font_setvaraxes #6690 [@cgohlke]

Only use ASCII characters in C source file #6691 [@cgohlke]

Release Python GIL when converting images using matrix operations #6418 [@hmaarrfk]

Added ExifTags enums #6630 [@radarhere]

Do not modify previous frame when calculating delta in PNG #6683 [@radarhere]

Added support for reading BMP images with RLE4 compression #6674 [@npjg]

Decode JPEG compressed BLP1 data in original mode #6678 [@radarhere]

pylint warnings #6659 [@marksmayo]

Added GPS TIFF tag info #6661 [@radarhere]

Added conversion between RGB/RGBA/RGBX and LAB #6647 [@radarhere]

Do not attempt normalization if mode is already normal #6644 [@radarhere]

Fixed seeking to an L frame in a GIF #6576 [@radarhere]

Consider all frames when selecting mode for PNG save_all #6610 [@radarhere]

Don't reassign crc on ChunkStream close #6627 [@radarhere]

Raise a warning if NumPy failed to raise an error during conversion #6594 [@radarhere]

Only read a maximum of 100 bytes at a time in IMT header #6623 [@radarhere]

Show all frames in ImageShow #6611 [@radarhere]

Allow FLI palette chunk to not be first #6626 [@radarhere]

If first GIF frame has transparency for RGB_ALWAYS loading strategy, use RGBA mode #6592 [@radarhere]

Round box position to integer when pasting embedded color #6517 [@radarhere]

Removed EXIF prefix when saving WebP #6582 [@radarhere]

Pad IM palette to 768 bytes when saving #6579 [@radarhere]

Added DDS BC6H reading #6449 [@ShadelessFox]

Added support for opening WhiteIsZero 16-bit integer TIFF images #6642 [@JayWiz]

Raise an error when allocating translucent color to RGB palette #6654 [@jsbueno]

Moved mode check outside of loops #6650 [@radarhere]

Added reading of TIFF child images #6569 [@radarhere]

Improved ImageOps palette handling #6596 [@PososikTeam]

Defer parsing of palette into colors #6567 [@radarhere]

Apply transparency to P images in ImageTk.PhotoImage #6559 [@radarhere]

Use rounding in ImageOps contain() and pad() #6522 [@bibinhashley]

Fixed GIF remapping to palette with duplicate entries #6548 [@radarhere]

Allow remap_palette() to return an image with less than 256 palette entries #6543 [@radarhere]

Corrected BMP and TGA palette size when saving #6500 [@radarhere]

... (truncated)

Changelog

Sourced from pillow's changelog.

9.3.0 (2022-10-29)

Limit SAMPLESPERPIXEL to avoid runtime DOS #6700 [wiredfool]

Initialize libtiff buffer when saving #6699 [radarhere]

Inline fname2char to fix memory leak #6329 [nulano]

Fix memory leaks related to text features #6330 [nulano]

Use double quotes for version check on old CPython on Windows #6695 [hugovk]

Remove backup implementation of Round for Windows platforms #6693 [cgohlke]

Fixed set_variation_by_name offset #6445 [radarhere]

Fix malloc in _imagingft.c:font_setvaraxes #6690 [cgohlke]

Release Python GIL when converting images using matrix operations #6418 [hmaarrfk]

Added ExifTags enums #6630 [radarhere]

Do not modify previous frame when calculating delta in PNG #6683 [radarhere]

Added support for reading BMP images with RLE4 compression #6674 [npjg, radarhere]

Decode JPEG compressed BLP1 data in original mode #6678 [radarhere]

Added GPS TIFF tag info #6661 [radarhere]

Added conversion between RGB/RGBA/RGBX and LAB #6647 [radarhere]

Do not attempt normalization if mode is already normal #6644 [radarhere]

... (truncated)

Commits

d594f4c Update CHANGES.rst [ci skip]

909dc64 9.3.0 version bump

1a51ce7 Merge pull request #6699 from hugovk/security-libtiff_buffer

2444cdd Merge pull request #6700 from hugovk/security-samples_per_pixel-sec

744f455 Added release notes

0846bfa Add to release notes

799a6a0 Fix linting

00b25fd Hide UserWarning in logs

05b175e Tighter test case

13f2c5a Prevent DOS with large SAMPLESPERPIXEL in Tiff IFD

Additional commits viewable in compare view

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

@dependabot rebase will rebase this PR

@dependabot recreate will recreate this PR, overwriting any edits that have been made to it

@dependabot merge will merge this PR after your CI passes on it

@dependabot squash and merge will squash and merge this PR after your CI passes on it

@dependabot cancel merge will cancel a previously requested merge and block automerging

@dependabot reopen will reopen this PR if it is closed

@dependabot close will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually

@dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)

@dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)

@dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)

@dependabot use these labels will set the current labels as the default for future PRs for this repo and language

@dependabot use these reviewers will set the current reviewers as the default for future PRs for this repo and language

@dependabot use these assignees will set the current assignees as the default for future PRs for this repo and language

@dependabot use this milestone will set the current milestone as the default for future PRs for this repo and language

You can disable automated security fix PRs for this repo from the Security Alerts page.

dependencies
opened by dependabot[bot] 0
Bump tensorflow from 2.5.1 to 2.9.3
Bumps tensorflow from 2.5.1 to 2.9.3.

Release notes

Sourced from tensorflow's releases.

TensorFlow 2.9.3

Release 2.9.3

This release introduces several vulnerability fixes:

Fixes an overflow in tf.keras.losses.poisson (CVE-2022-41887)

Fixes a heap OOB failure in ThreadUnsafeUnigramCandidateSampler caused by missing validation (CVE-2022-41880)

Fixes a segfault in ndarray_tensor_bridge (CVE-2022-41884)

Fixes an overflow in FusedResizeAndPadConv2D (CVE-2022-41885)

Fixes a overflow in ImageProjectiveTransformV2 (CVE-2022-41886)

Fixes an FPE in tf.image.generate_bounding_box_proposals on GPU (CVE-2022-41888)

Fixes a segfault in pywrap_tfe_src caused by invalid attributes (CVE-2022-41889)

Fixes a CHECK fail in BCast (CVE-2022-41890)

Fixes a segfault in TensorListConcat (CVE-2022-41891)

Fixes a CHECK_EQ fail in TensorListResize (CVE-2022-41893)

Fixes an overflow in CONV_3D_TRANSPOSE on TFLite (CVE-2022-41894)

Fixes a heap OOB in MirrorPadGrad (CVE-2022-41895)

Fixes a crash in Mfcc (CVE-2022-41896)

Fixes a heap OOB in FractionalMaxPoolGrad (CVE-2022-41897)

Fixes a CHECK fail in SparseFillEmptyRowsGrad (CVE-2022-41898)

Fixes a CHECK fail in SdcaOptimizer (CVE-2022-41899)

Fixes a heap OOB in FractionalAvgPool and FractionalMaxPool(CVE-2022-41900)

Fixes a CHECK_EQ in SparseMatrixNNZ (CVE-2022-41901)

Fixes an OOB write in grappler (CVE-2022-41902)

Fixes a overflow in ResizeNearestNeighborGrad (CVE-2022-41907)

Fixes a CHECK fail in PyFunc (CVE-2022-41908)

Fixes a segfault in CompositeTensorVariantToComponents (CVE-2022-41909)

Fixes a invalid char to bool conversion in printing a tensor (CVE-2022-41911)

Fixes a heap overflow in QuantizeAndDequantizeV2 (CVE-2022-41910)

Fixes a CHECK failure in SobolSample via missing validation (CVE-2022-35935)

Fixes a CHECK fail in TensorListScatter and TensorListScatterV2 in eager mode (CVE-2022-35935)

TensorFlow 2.9.2

Release 2.9.2

This releases introduces several vulnerability fixes:

Fixes a CHECK failure in tf.reshape caused by overflows (CVE-2022-35934)

Fixes a CHECK failure in SobolSample caused by missing validation (CVE-2022-35935)

Fixes an OOB read in Gather_nd op in TF Lite (CVE-2022-35937)

Fixes a CHECK failure in TensorListReserve caused by missing validation (CVE-2022-35960)

Fixes an OOB write in Scatter_nd op in TF Lite (CVE-2022-35939)

Fixes an integer overflow in RaggedRangeOp (CVE-2022-35940)

Fixes a CHECK failure in AvgPoolOp (CVE-2022-35941)

Fixes a CHECK failures in UnbatchGradOp (CVE-2022-35952)

Fixes a segfault TFLite converter on per-channel quantized transposed convolutions (CVE-2022-36027)

Fixes a CHECK failures in AvgPool3DGrad (CVE-2022-35959)

Fixes a CHECK failures in FractionalAvgPoolGrad (CVE-2022-35963)

Fixes a segfault in BlockLSTMGradV2 (CVE-2022-35964)

Fixes a segfault in LowerBound and UpperBound (CVE-2022-35965)

... (truncated)

Changelog

Sourced from tensorflow's changelog.

Release 2.9.3

This release introduces several vulnerability fixes:

Fixes an overflow in tf.keras.losses.poisson (CVE-2022-41887)

Fixes a heap OOB failure in ThreadUnsafeUnigramCandidateSampler caused by missing validation (CVE-2022-41880)

Fixes a segfault in ndarray_tensor_bridge (CVE-2022-41884)

Fixes an overflow in FusedResizeAndPadConv2D (CVE-2022-41885)

Fixes a overflow in ImageProjectiveTransformV2 (CVE-2022-41886)

Fixes an FPE in tf.image.generate_bounding_box_proposals on GPU (CVE-2022-41888)

Fixes a segfault in pywrap_tfe_src caused by invalid attributes (CVE-2022-41889)

Fixes a CHECK fail in BCast (CVE-2022-41890)

Fixes a segfault in TensorListConcat (CVE-2022-41891)

Fixes a CHECK_EQ fail in TensorListResize (CVE-2022-41893)

Fixes an overflow in CONV_3D_TRANSPOSE on TFLite (CVE-2022-41894)

Fixes a heap OOB in MirrorPadGrad (CVE-2022-41895)

Fixes a crash in Mfcc (CVE-2022-41896)

Fixes a heap OOB in FractionalMaxPoolGrad (CVE-2022-41897)

Fixes a CHECK fail in SparseFillEmptyRowsGrad (CVE-2022-41898)

Fixes a CHECK fail in SdcaOptimizer (CVE-2022-41899)

Fixes a heap OOB in FractionalAvgPool and FractionalMaxPool(CVE-2022-41900)

Fixes a CHECK_EQ in SparseMatrixNNZ (CVE-2022-41901)

Fixes an OOB write in grappler (CVE-2022-41902)

Fixes a overflow in ResizeNearestNeighborGrad (CVE-2022-41907)

Fixes a CHECK fail in PyFunc (CVE-2022-41908)

Fixes a segfault in CompositeTensorVariantToComponents (CVE-2022-41909)

Fixes a invalid char to bool conversion in printing a tensor (CVE-2022-41911)

Fixes a heap overflow in QuantizeAndDequantizeV2 (CVE-2022-41910)

Fixes a CHECK failure in SobolSample via missing validation (CVE-2022-35935)

Fixes a CHECK fail in TensorListScatter and TensorListScatterV2 in eager mode (CVE-2022-35935)

Release 2.8.4

This release introduces several vulnerability fixes:

Fixes a heap OOB failure in ThreadUnsafeUnigramCandidateSampler caused by missing validation (CVE-2022-41880)

Fixes a segfault in ndarray_tensor_bridge (CVE-2022-41884)

Fixes an overflow in FusedResizeAndPadConv2D (CVE-2022-41885)

Fixes a overflow in ImageProjectiveTransformV2 (CVE-2022-41886)

Fixes an FPE in tf.image.generate_bounding_box_proposals on GPU (CVE-2022-41888)

Fixes a segfault in pywrap_tfe_src caused by invalid attributes (CVE-2022-41889)

Fixes a CHECK fail in BCast (CVE-2022-41890)

Fixes a segfault in TensorListConcat (CVE-2022-41891)

Fixes a CHECK_EQ fail in TensorListResize (CVE-2022-41893)

Fixes an overflow in CONV_3D_TRANSPOSE on TFLite (CVE-2022-41894)

Fixes a heap OOB in MirrorPadGrad (CVE-2022-41895)

Fixes a crash in Mfcc (CVE-2022-41896)

Fixes a heap OOB in FractionalMaxPoolGrad (CVE-2022-41897)

Fixes a CHECK fail in SparseFillEmptyRowsGrad (CVE-2022-41898)

Fixes a CHECK fail in SdcaOptimizer (CVE-2022-41899)

... (truncated)

Commits

a5ed5f3 Merge pull request #58584 from tensorflow/vinila21-patch-2

258f9a1 Update py_func.cc

cd27cfb Merge pull request #58580 from tensorflow-jenkins/version-numbers-2.9.3-24474

3e75385 Update version numbers to 2.9.3

bc72c39 Merge pull request #58482 from tensorflow-jenkins/relnotes-2.9.3-25695

3506c90 Update RELEASE.md

8dcb48e Update RELEASE.md

4f34ec8 Merge pull request #58576 from pak-laura/c2.99f03a9d3bafe902c1e6beb105b2f2417...

6fc67e4 Replace CHECK with returning an InternalError on failing to create python tuple

5dbe90a Merge pull request #58570 from tensorflow/r2.9-7b174a0f2e4

Additional commits viewable in compare view

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

@dependabot rebase will rebase this PR

@dependabot recreate will recreate this PR, overwriting any edits that have been made to it

@dependabot merge will merge this PR after your CI passes on it

@dependabot squash and merge will squash and merge this PR after your CI passes on it

@dependabot cancel merge will cancel a previously requested merge and block automerging

@dependabot reopen will reopen this PR if it is closed

@dependabot close will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually

@dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)

@dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)

@dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)

@dependabot use these labels will set the current labels as the default for future PRs for this repo and language

@dependabot use these reviewers will set the current reviewers as the default for future PRs for this repo and language

@dependabot use these assignees will set the current assignees as the default for future PRs for this repo and language

@dependabot use this milestone will set the current milestone as the default for future PRs for this repo and language

You can disable automated security fix PRs for this repo from the Security Alerts page.

dependencies
opened by dependabot[bot] 0
Reproduce Fig.4. Zebra TCAV in googlenet from paper

I was wondering if it should be possible to reproduce the results from fig.4. from the TCAV-paper (Zebra TCAV in googlenet) using the NB tcav/Run_TCAV.ipynb?

opened by gaunby 0
Bump protobuf from 3.10.0 to 3.18.3
Bumps protobuf from 3.10.0 to 3.18.3.

Release notes

Sourced from protobuf's releases.

Protocol Buffers v3.18.3

C++

Reduce memory consumption of MessageSet parsing

This release addresses a Security Advisory for C++ and Python users

Protocol Buffers v3.16.1

Java

Improve performance characteristics of UnknownFieldSet parsing (#9371)

Protocol Buffers v3.18.2

Java

Improve performance characteristics of UnknownFieldSet parsing (#9371)

Protocol Buffers v3.18.1

Python

Update setup.py to reflect that we now require at least Python 3.5 (#8989)

Performance fix for DynamicMessage: force GetRaw() to be inlined (#9023)

Ruby

Update ruby_generator.cc to allow proto2 imports in proto3 (#9003)

Protocol Buffers v3.18.0

C++

Fix warnings raised by clang 11 (#8664)

Make StringPiece constructible from std::string_view (#8707)

Add missing capability attributes for LLVM 12 (#8714)

Stop using std::iterator (deprecated in C++17). (#8741)

Move field_access_listener from libprotobuf-lite to libprotobuf (#8775)

Fix #7047 Safely handle setlocale (#8735)

Remove deprecated version of SetTotalBytesLimit() (#8794)

Support arena allocation of google::protobuf::AnyMetadata (#8758)

Fix undefined symbol error around SharedCtor() (#8827)

Fix default value of enum(int) in json_util with proto2 (#8835)

Better Smaller ByteSizeLong

Introduce event filters for inject_field_listener_events

Reduce memory usage of DescriptorPool

For lazy fields copy serialized form when allowed.

Re-introduce the InlinedStringField class

v2 access listener

Reduce padding in the proto's ExtensionRegistry map.

GetExtension performance optimizations

Make tracker a static variable rather than call static functions

Support extensions in field access listener

Annotate MergeFrom for field access listener

Fix incomplete types for field access listener

Add map_entry/new_map_entry to SpecificField in MessageDifferencer. They record the map items which are different in MessageDifferencer's reporter.

Reduce binary size due to fieldless proto messages

TextFormat: ParseInfoTree supports getting field end location in addition to start.

... (truncated)

Commits

a902b39 No-op whitespace change

ae62acd Updating version.json and repo version numbers to: 18.3

f43ac49 Merge pull request #10542 from deannagarcia/3.18.x

9efdf55 Add missing includes

d1635e1 Apply patch

5b37c91 Update version.json with "lts": true (#10534)

c39d622 Merge pull request #10529 from protocolbuffers/deannagarcia-patch-5

f77d3b6 Update version.json

8178b06 Merge pull request #10503 from deannagarcia/3.18.x

24ca839 Add version file

Additional commits viewable in compare view

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

@dependabot rebase will rebase this PR

@dependabot recreate will recreate this PR, overwriting any edits that have been made to it

@dependabot merge will merge this PR after your CI passes on it

@dependabot squash and merge will squash and merge this PR after your CI passes on it

@dependabot cancel merge will cancel a previously requested merge and block automerging

@dependabot reopen will reopen this PR if it is closed

@dependabot close will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually

@dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)

@dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)

@dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)

@dependabot use these labels will set the current labels as the default for future PRs for this repo and language

@dependabot use these reviewers will set the current reviewers as the default for future PRs for this repo and language

@dependabot use these assignees will set the current assignees as the default for future PRs for this repo and language

@dependabot use this milestone will set the current milestone as the default for future PRs for this repo and language

You can disable automated security fix PRs for this repo from the Security Alerts page.

dependencies
opened by dependabot[bot] 0

Releases(0.2)

0.2(Nov 21, 2018)

This release has an API redesign that uses an ActivationGenerator class instead of the old tcav_helpers file. This simplies using TCAV with all different types of models with all different types of examples and ways of loading examples.
Source code(tar.gz)
Source code(zip)
0.1(Nov 14, 2018)

Initial TCAV code before refactor
Source code(tar.gz)
Source code(zip)

Owner

GitHub Repository

A mindmap summarising Machine Learning concepts, from Data Analysis to Deep Learning.

5.7k Dec 30, 2022

An MLOps framework to package, deploy, monitor and manage thousands of production machine learning models

Seldon Core: Blazing Fast, Industry-Ready ML An open source platform to deploy your machine learning models on Kubernetes at massive scale. Overview S

3.5k Jan 01, 2023

Forecasting prices using Facebook/Meta's Prophet model

CryptoForecasting using Machine and Deep learning (Part 1) CryptoForecasting using Machine Learning The main aspect of predicting the stock-related da

1 Nov 27, 2021

Predicting Baseball Metric Clusters: Clustering Application in Python Using scikit-learn

Clustering Clustering Application in Python Using scikit-learn This repository contains the prediction of baseball metric clusters using MLB Statcast

2 Apr 18, 2022

Python library for multilinear algebra and tensor factorizations

scikit-tensor is a Python module for multilinear algebra and tensor factorizations

394 Dec 09, 2022

This is a Cricket Score Predictor that predicts the first innings score of a T20 Cricket match using Machine Learning

This is a Cricket Score Predictor that predicts the first innings score of a T20 Cricket match using Machine Learning. It is a Web Application.

3 Aug 04, 2022

Project to deploy a machine learning model based on Titanic dataset from Kaggle

kaggle_titanic_deploy Project to deploy a machine learning model based on Titanic dataset from Kaggle In this project we used the Titanic dataset from

8 May 23, 2022

This repository has datasets containing information of Uber pickups in NYC from April 2014 to September 2014 and January to June 2015. data Analysis , virtualization and some insights are gathered here

uber-pickups-analysis Data Source: https://www.kaggle.com/fivethirtyeight/uber-pickups-in-new-york-city Information about data set The dataset contain

1 Nov 03, 2021

50% faster, 50% less RAM Machine Learning. Numba rewritten Sklearn. SVD, NNMF, PCA, LinearReg, RidgeReg, Randomized, Truncated SVD/PCA, CSR Matrices all 50+% faster

[Due to the time taken @ uni, work + hell breaking loose in my life, since things have calmed down a bit, will continue commiting!!!] [By the way, I'm

1.4k Jan 01, 2023

A Python step-by-step primer for Machine Learning and Optimization

early-ML Presentation General Machine Learning tutorials A Python step-by-step primer for Machine Learning and Optimization This github repository gat

8 Dec 01, 2022

A Python library for choreographing your machine learning research.

270 Jan 06, 2023

To-Be is a machine learning challenge on CodaLab Platform about Mortality Prediction

To-Be is a machine learning challenge on CodaLab Platform about Mortality Prediction. The challenge aims to adress the problems of medical imbalanced data classification.

1 Jan 31, 2022

Little Ball of Fur - A graph sampling extension library for NetworKit and NetworkX (CIKM 2020)

Little Ball of Fur is a graph sampling extension library for Python. Please look at the Documentation, relevant Paper, Promo video and External Resour

619 Dec 14, 2022

CyLP is a Python interface to COIN-OR’s Linear and mixed-integer program solvers (CLP, CBC, and CGL)

CyLP CyLP is a Python interface to COIN-OR’s Linear and mixed-integer program solvers (CLP, CBC, and CGL). CyLP’s unique feature is that you can use i

161 Dec 14, 2022

Databricks Certified Associate Spark Developer preparation toolkit to setup single node Standalone Spark Cluster along with material in the form of Jupyter Notebooks.

Databricks Certification Spark Databricks Certified Associate Spark Developer preparation toolkit to setup single node Standalone Spark Cluster along

19 Dec 13, 2022

A Lightweight Hyperparameter Optimization Tool 🚀

The mle-hyperopt package provides a simple and intuitive API for hyperparameter optimization of your Machine Learning Experiment (MLE) pipeline.

137 Dec 02, 2022

moDel Agnostic Language for Exploration and eXplanation

moDel Agnostic Language for Exploration and eXplanation Overview Unverified black box model is the path to the failure. Opaqueness leads to distrust.

1.2k Jan 04, 2023

nn-Meter is a novel and efficient system to accurately predict the inference latency of DNN models on diverse edge devices

A DNN inference latency prediction toolkit for accurately modeling and predicting the latency on diverse edge devices.

241 Dec 26, 2022

cleanlab is the data-centric ML ops package for machine learning with noisy labels.

cleanlab is the data-centric ML ops package for machine learning with noisy labels. cleanlab cleans labels and supports finding, quantifying, and lear

51 Nov 28, 2022

A chain of stores, 10 different stores and 50 different requests a 3-month demand forecast for its product.

Demand-Forecasting Business Problem A chain of stores, 10 different stores and 50 different requests a 3-month demand forecast for its product.

3 Mar 06, 2022

Code for the TCAV ML interpretability project

Related tags

Overview

Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)

What is TCAV?

What's special about TCAV compared to other methods?

Cool, where do these concepts come from?

Why use high level concepts instead of input features?

The consumer of the explanation may not know machine learning too well. Can they understand the explanation?

Sounds good. Do I need to change my network to use TCAV?

Installation

Requirements

How to use TCAV

TCAV for discrete models

Requirements

How to run unit tests

How to create a new version of the pip package

Comments

Change list

9.3.0

Changes

9.3.0 (2022-10-29)

TensorFlow 2.9.3

Release 2.9.3

TensorFlow 2.9.2

Release 2.9.2

Release 2.9.3

Release 2.8.4

Protocol Buffers v3.18.3

C++

Protocol Buffers v3.16.1

Java

Protocol Buffers v3.18.2

Java

Protocol Buffers v3.18.1

Python

Ruby

Protocol Buffers v3.18.0

C++

Releases(0.2)

0.2(Nov 21, 2018)

0.1(Nov 14, 2018)

Owner

A mindmap summarising Machine Learning concepts, from Data Analysis to Deep Learning.

An MLOps framework to package, deploy, monitor and manage thousands of production machine learning models

Forecasting prices using Facebook/Meta's Prophet model

Predicting Baseball Metric Clusters: Clustering Application in Python Using scikit-learn

Python library for multilinear algebra and tensor factorizations

This is a Cricket Score Predictor that predicts the first innings score of a T20 Cricket match using Machine Learning

Project to deploy a machine learning model based on Titanic dataset from Kaggle

This repository has datasets containing information of Uber pickups in NYC from April 2014 to September 2014 and January to June 2015. data Analysis , virtualization and some insights are gathered here

50% faster, 50% less RAM Machine Learning. Numba rewritten Sklearn. SVD, NNMF, PCA, LinearReg, RidgeReg, Randomized, Truncated SVD/PCA, CSR Matrices all 50+% faster

A Python step-by-step primer for Machine Learning and Optimization

A Python library for choreographing your machine learning research.

To-Be is a machine learning challenge on CodaLab Platform about Mortality Prediction

Little Ball of Fur - A graph sampling extension library for NetworKit and NetworkX (CIKM 2020)

CyLP is a Python interface to COIN-OR’s Linear and mixed-integer program solvers (CLP, CBC, and CGL)

Databricks Certified Associate Spark Developer preparation toolkit to setup single node Standalone Spark Cluster along with material in the form of Jupyter Notebooks.

A Lightweight Hyperparameter Optimization Tool 🚀

moDel Agnostic Language for Exploration and eXplanation

nn-Meter is a novel and efficient system to accurately predict the inference latency of DNN models on diverse edge devices

cleanlab is the data-centric ML ops package for machine learning with noisy labels.

A chain of stores, 10 different stores and 50 different requests a 3-month demand forecast for its product.