Convolutional Neural Network for auto-tagging of audio clips on MagnaTagATune dataset

emarkou, updated 🕥 2022-08-23 18:16:27

Development of a Convolutional Neural Network for multi label auto tagging of music audio files

HitCount

Initial steps

Download mp3 files and combine them using: cat mp3.zip.* > single_mp3.zip Extract files from subfolders: find . -mindepth 2 -type f -print -exec mv {} . \;

Introduction

In general, music audio files can be accompanied by metadata related to their content such as free text description or tags. Tags are proved to be more useful as they provide a more direct description of the audio file and can be used in tasks such as classification per gender, artist, musical instrument etc in recommendation systems related to music. As not all audio files are accompanied by tags, the need of auto-tagging arises.

One approach widely used involves the usage of unsupervised feature learning such as K-means, sparse coding and Boltzmann machines. In these cases, the main concern is the capture of low leve music structures that can be used as input into some classifier.

Another approach involves supervised methods, like Deep Neural Networks (DNNs) of various architectural types (MLP, CNN, RNN), that directly map labels to audio files. In these cases, the feature extraction method could vary a lot, from spectograms to hand-engineered features like Mel Frequency Cepstral Coefficients (MFCCs).

In more detail, a spectrogram of an audio clip is a visual representation of its spectrum of frequencies as they vary with time. A common variation is the mel spectrogram, where the so-called mel scale, i.e. a perceptual scale of pitches judged by listeners to be equal in distance from one another, is used. Additionally, MFCCs are coefficients that collectively make up a Mel Frequency Cepstrum (MFC), which is a representation of the short-term power spectrum of a asound, based on a linear cosine transform of a log power spectrum on a nonlinear mel scale of frequency.

For the purpose of this project, the method that was deployed involves the transformation of the initial sound files into 2D numerical matrices using both feature extraction techniques (mel spectrograms, MFCCs). These matrices were then fed into a CNN in order to assign suitable tags to each audio clip. Multiple architectures, from shallow to deeper, were evaluated.

The dataset

Among many candidates, the one finally selected was the MagnaTagATune dataset which can be found here. The dataset consists of almost 26000 audio tracks of 29 seconds in mp3 format with 32 mbps, where the 188 available binary tags are not mutually exclusive (i.e. a file can be labeled with more than one tag). The dataset was preferred among all others, since:

  • All clips have exactly the same duration
  • All tags are organized in a matrix format with reference to each clip via a unique clip id
  • This dataset has already been used successfully in many academic publications
  • The dataset is free of copyright protection (DRM licensing) as part of the magnatune music collection

The available tags include, among others music genres (e.g. "pop", "alternative", "indie"), instruments (e.g. "guitar", "violin", "piano"), emotions (e.g. "mellow", "chill") and origins (e.g. "middle eastern", "India"). There are also some tags representing the absence of a characteristic (e.g. "no beat", "no drums")

Data preprocessing

Label reduction

The tags of the original dataset displayed two main dysfunctionalities: synonym tags and extremely skewed tag distributions, i.e. existence of tags with rare occurence. To solve the above issues, in 1_label_reduction_and_files_renaming.py and aux_label_distribution.py, synonym and similar tags were merged into a single tage and eventually the 50 most popular of the merged tags were kept.

Feature extraction

For each audio file, the below process was applied: - Selection of the desirable sampling rate. Uniform sample rates were used to allow easy batching of the data - Selection of the desirable number and type of features (mel spectrogram, MFCCs) to be extracted per sampling window. - Extraction of features from the .wav file, i.e. generation of a 2D array where the horizontal axis represents the sampling windows and the vertical axis represents the extracted features. - Storage of the above information in numpy array format.

The result of the process is the transformation of the audio wave into a 2D matrix of features. This is displayed in the following figure for three tracks where the frequencies are shown increasing up the vertical axis, time is depicted on the horizontal axis and wave amplitude is represented as color intensity. Sound waves, Spectrograms & log Spectrograms

Methodology

The general idea for the CNN construction was to use the generated 2D numpy arrays as input into a chain of two or three convolutional layers (with or without max pooling), followed by a fully connected layer prior to the class prediction output (multi-hot vectors of 50 positions, one for each tag).

  • Activation Function: ReLu
  • Loss Function: sigmoid cross entropy with logits
  • Optimizer: Adaptive Moment Estimation, i.e. Adam

The finnaly selected architecture is the following: CNN architecure

Results

As mentioned previously, the auto tagging task is a multi label classification problem. Thus, th ground truth labels provided along with the MagnaTagATune data are encoded as a 50x1 binary vector (multihot vector) where each dimension corresponds to a specific tag. Accordingly, the predictions are also represented as a vector of the same size, but in this case the values are not binary anymore; instead, the represent the probability of that label being assigned to the corresponding audio clip. By setting a threshold at 0.5 the final set of labels was acquired.

To assess the network's performance the Receiver Operating Characteristic Curve (AUC) was used which represents the probability that a random classifier will rank a randomly chosen positive instance higher that a randomly chosen negative one. Regarding the Accuracy and True Positive Rate per tag, it was observed that classes with few occurrences display a higher accuracy percentage. This happens due to the high value of True Negatives. In contrast, classes with frequent appearance, such as "guitar", "techno" and "classical", may have a lower accuracy percentage but they are the ones that the network has learnt better, since they have the highest True Positive Rate (Recall).

Accuracy

Recall

Visualizations

As mentioned before, the initial tagging of the clips included in the MagnaTagATune dataset is done by humans. Consequently, in many cases not all appropriate tags were assigned to the corresponding songs. Therefore, for the easier evaluation of the appropriateness of the assigned tags to each audio clip, a visual representation of the tags was constructed.

Words were represented as vectors with the use of a word to vector model. The word embeddings model was generated with Google News word to vector pre-trained model, comprised of 300 dimensions. The Google News Model is a skip-gram model with negative sampling that is proven to capture a large number of precise syntactic and semantic word relationships. Visualization of music labels in a 2D image required the dimensionality reduction from 300 dimensions to 2. Instead of the Principal Component Analysis (PCA) method, the Distributed Stochastic Neighbor Embedding method (t-SNE) was used. The t-SNE achieves dimensionality reduction by preserving, as much as possible, the original semantic similarity of the tags.

t-SNE visualization

It should be noted that the cost function used by the method is non-convex while the optimization algorithm is the gradient descent. Thus, different executions of the algorithm would lead to different results. To accomplish a more comprehensive and interactive visualization Tensorflow Projector was used. For this purpose, t-SNE algorithm was implemented twicel; once to reduce Google News model dimensions from 300 to 4 and a second time when Tensorflow Projector was used.

Studying the produced visualization, it becomes evident that the majority of the semantically close tags are clustered together. Some indicative examples of clusters include:

  • Jazz, slow, soft, quiet, piano, string
  • Drum, synth, dance, modern, electro,
  • Orchestra, ambient, harpsichord, harp, violin, flute

To observe more closely the results of the CNN some indicative examples are presented below. Results

It is identifies that the provided dataset annotations in specific tracks may be insufficient. For example, for clip id 15317 the annotation is "no piano" which provides no information regarding the song itself. This consequently makes training of the CNN more challenging.

Furthermore, it should be noted that the predicted labels in the majority of cases may be not be 100% accurate but are semantically close to the provided ones. For example, it is evident that in clip id 35334 the annotation "guitar" has been translated to "bass" and "violin" which are all stringed instruments.

In addition, it is observed that the model predicts suitable tags even in cases where they are not part of the initial annotations, as happens with clip id 24335.

Issues

Bump nbconvert from 5.3.1 to 6.5.1

opened on 2022-08-23 18:16:23 by dependabot[bot]

Bumps nbconvert from 5.3.1 to 6.5.1.

Release notes

Sourced from nbconvert's releases.

Release 6.5.1

No release notes provided.

6.5.0

What's Changed

New Contributors

Full Changelog: https://github.com/jupyter/nbconvert/compare/6.4.5...6.5

6.4.3

What's Changed

New Contributors

Full Changelog: https://github.com/jupyter/nbconvert/compare/6.4.2...6.4.3

6.4.0

What's Changed

New Contributors

... (truncated)

Commits


Dependabot compatibility score

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](https://github.com/emarkou/Audio-auto-tagging/network/alerts).

Bump pyarrow from 0.12.0 to 0.15.1

opened on 2022-08-05 21:58:13 by dependabot[bot]

Bumps pyarrow from 0.12.0 to 0.15.1.

Changelog

Sourced from pyarrow's changelog.

Apache Arrow 0.15.1 (25 October 2019)

Bug

  • ARROW-6464 - [Java] Refactor FixedSizeListVector#splitAndTransfer with slice API
  • ARROW-6728 - [C#] Support reading and writing Date32 and Date64 arrays
  • ARROW-6740 - [Python] Unable to delete closed MemoryMappedFile on Windows
  • ARROW-6762 - [C++] JSON reader segfaults on newline
  • ARROW-6795 - [C#] Reading large Arrow files in C# results in an exception
  • ARROW-6806 - [C++] Segfault deserializing ListArray containing null/empty list
  • ARROW-6813 - [Ruby] Arrow::Table.load with headers=true leads to exception in Arrow 0.15
  • ARROW-6834 - [C++] Pin gtest to 1.8.1 to triage failing Appveyor / MSVC build
  • ARROW-6844 - [C++][Parquet][Python] List columns read broken with 0.15.0
  • ARROW-6857 - [Python][C++] Segfault for dictionary_encode on empty chunked_array (edge case)
  • ARROW-6860 - [Python] Only link libarrow_flight.so to pyarrow._flight
  • ARROW-6861 - [Python] arrow-0.15.0 reading arrow-0.14.1-output Parquet dictionary column: Failure reading column: IOError: Arrow error: Invalid: Resize cannot downsize
  • ARROW-6869 - [C++] Dictionary "delta" building logic in builder_dict.h produces invalid arrays
  • ARROW-6873 - [Python] Stale CColumn reference break Cython cimport pyarrow
  • ARROW-6874 - [Python] Memory leak in Table.to_pandas() when conversion to object dtype
  • ARROW-6876 - [Python] Reading parquet file with many columns becomes slow for 0.15.0
  • ARROW-6877 - [C++] Boost not found from the correct environment
  • ARROW-6878 - [Python] pa.array() does not handle list of dicts with bytes keys correctly under python3
  • ARROW-6882 - [Python] cannot create a chunked_array from dictionary_encoding result
  • ARROW-6886 - [C++] arrow::io header nvcc compiler warnings
  • ARROW-6898 - [Java] Fix potential memory leak in ArrowWriter and several test classes
  • ARROW-6903 - [Python] Wheels broken after ARROW-6860 changes
  • ARROW-6905 - [Packaging][OSX] Nightly builds on MacOS are failing because of brew compile timeouts
  • ARROW-6910 - [Python] pyarrow.parquet.read_table(...) takes up lots of memory which is not released until program exits
  • ARROW-6922 - [Python] Pandas master build is failing (MultiIndex.levels change)
  • ARROW-6937 - [Packaging][Python] Fix conda linux and OSX wheel nightly builds
  • ARROW-6938 - [Python] Windows wheel depends on zstd.dll and libbz2.dll, which are not bundled
  • ARROW-6962 - [C++] [CI] Stop compiling with -Weverything
  • ARROW-6977 - [C++] Only enable jemalloc background_thread if feature is supported
  • ARROW-6983 - [C++] Threaded task group crashes sometimes

Improvement

  • ARROW-6610 - [C++] Add ARROW_FILESYSTEM=ON/OFF CMake configuration flag
  • ARROW-6777 - [GLib][CI] Unpin gobject-introspection gem
  • ARROW-6852 - [C++] memory-benchmark build failed on Arm64
  • ARROW-6927 - [C++] Add gRPC version check
  • ARROW-6963 - [Packaging][Wheel][OSX] Use crossbow's command to deploy artifacts from travis builds

New Feature

  • ARROW-6661 - [Java] Implement APIs like slice to enhance VectorSchemaRoot

Apache Arrow 0.15.0 (30 September 2019)

Bug

... (truncated)

Commits
  • b789226 [maven-release-plugin] prepare release apache-arrow-0.15.1
  • 6687aed [Release] Set Maven versions to 0.15.1-SNAPSHOT
  • e9f7864 [Release] Update versions for 0.15.1
  • c7e9fe4 [Release] Update .deb/.rpm changelogs for 0.15.1
  • 6a94d72 [Release] Update CHANGELOG.md for 0.15.1
  • d02f74e ARROW-6962: [C++] [CI] Stop compiling with -Weverything
  • 28d6d97 ARROW-6963: [Packaging][Wheel][OSX] Use crossbow's command to deploy artifact...
  • a83a0f0 ARROW-6983: [C++] Fix ThreadedTaskGroup lifetime issue
  • 4142ed5 ARROW-6977: [C++] Disable jemalloc background_thread on macOS
  • c20eceb ARROW-6910: [C++][Python] Set jemalloc default configuration to release dirty...
  • Additional commits viewable in compare view


Dependabot compatibility score

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](https://github.com/emarkou/Audio-auto-tagging/network/alerts).

Bump mistune from 0.8.3 to 2.0.3

opened on 2022-07-29 23:14:51 by dependabot[bot]

Bumps mistune from 0.8.3 to 2.0.3.

Release notes

Sourced from mistune's releases.

Version 2.0.2

Fix escape_url via lepture/mistune#295

Version 2.0.1

Fix XSS for image link syntax.

Version 2.0.0

First release of Mistune v2.

Version 2.0.0 RC1

In this release, we have a Security Fix for harmful links.

Version 2.0.0 Alpha 1

This is the first release of v2. An alpha version for users to have a preview of the new mistune.

Version 0.8.4

  • Support an escaped pipe char in a table cell. #150
  • Fix ordered and unordered list. #152
  • Fix spaces between = in HTML tags
  • Add max_recursive_depth for list and blockquote.
  • Fix fences code block.
Changelog

Sourced from mistune's changelog.

Changelog

Here is the full history of mistune v2.

Version 2.0.4


Released on Jul 15, 2022

    

  • Fix url plugin in <a> tag
    • 

  • Fix * formatting
    • 



Version 2.0.3

Released on Jun 27, 2022

  • Fix table plugin
  • Security fix for CVE-2022-34749

Version 2.0.2


Released on Jan 14, 2022

Fix escape_url


Version 2.0.1

Released on Dec 30, 2021

XSS fix for image link syntax.

Version 2.0.0


Released on Dec 5, 2021

This is the first non-alpha release of mistune v2.


Version 2.0.0rc1

Released on Feb 16, 2021

Version 2.0.0a6


</tr></table>

... (truncated)

Commits
  • 3f422f1 Version bump 2.0.3
  • a6d4321 Fix asteris emphasis regex CVE-2022-34749
  • 5638e46 Merge pull request #307 from jieter/patch-1
  • 0eba471 Fix typo in guide.rst
  • 61e9337 Fix table plugin
  • 76dec68 Add documentation for renderer heading when TOC enabled
  • 799cd11 Version bump 2.0.2
  • babb0cf Merge pull request #295 from dairiki/bug.escape_url
  • fc2cd53 Make mistune.util.escape_url less aggressive
  • 3e8d352 Version bump 2.0.1
  • Additional commits viewable in compare view


Dependabot compatibility score

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](https://github.com/emarkou/Audio-auto-tagging/network/alerts).

Bump distributed from 1.21.8 to 2021.10.0

opened on 2022-07-15 22:00:42 by dependabot[bot]

Bumps distributed from 1.21.8 to 2021.10.0.

Commits


Dependabot compatibility score

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](https://github.com/emarkou/Audio-auto-tagging/network/alerts).

Bump lxml from 4.6.3 to 4.9.1

opened on 2022-07-06 20:51:26 by dependabot[bot]

Bumps lxml from 4.6.3 to 4.9.1.

Changelog

Sourced from lxml's changelog.

4.9.1 (2022-07-01)

Bugs fixed

  • A crash was resolved when using iterwalk() (or canonicalize()) after parsing certain incorrect input. Note that iterwalk() can crash on valid input parsed with the same parser after failing to parse the incorrect input.

4.9.0 (2022-06-01)

Bugs fixed

  • GH#341: The mixin inheritance order in lxml.html was corrected. Patch by xmo-odoo.

Other changes

  • Built with Cython 0.29.30 to adapt to changes in Python 3.11 and 3.12.

  • Wheels include zlib 1.2.12, libxml2 2.9.14 and libxslt 1.1.35 (libxml2 2.9.12+ and libxslt 1.1.34 on Windows).

  • GH#343: Windows-AArch64 build support in Visual Studio. Patch by Steve Dower.

4.8.0 (2022-02-17)

Features added

  • GH#337: Path-like objects are now supported throughout the API instead of just strings. Patch by Henning Janssen.

  • The ElementMaker now supports QName values as tags, which always override the default namespace of the factory.

Bugs fixed

  • GH#338: In lxml.objectify, the XSI float annotation "nan" and "inf" were spelled in lower case, whereas XML Schema datatypes define them as "NaN" and "INF" respectively.

... (truncated)

Commits
  • d01872c Prevent parse failure in new test from leaking into later test runs.
  • d65e632 Prepare release of lxml 4.9.1.
  • 86368e9 Fix a crash when incorrect parser input occurs together with usages of iterwa...
  • 50c2764 Delete unused Travis CI config and reference in docs (GH-345)
  • 8f0bf2d Try to speed up the musllinux AArch64 build by splitting the different CPytho...
  • b9f7074 Remove debug print from test.
  • b224e0f Try to install 'xz' in wheel builds, if available, since it's now needed to e...
  • 897ebfa Update macOS deployment target version from 10.14 to 10.15 since 10.14 starts...
  • 853c9e9 Prepare release of 4.9.0.
  • d3f77e6 Add a test for https://bugs.launchpad.net/lxml/+bug/1965070 leaving out the a...
  • Additional commits viewable in compare view


Dependabot compatibility score

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](https://github.com/emarkou/Audio-auto-tagging/network/alerts).

Bump numpy from 1.16.6 to 1.22.0

opened on 2022-06-22 03:35:27 by dependabot[bot]

Bumps numpy from 1.16.6 to 1.22.0.

Release notes

Sourced from numpy's releases.

v1.22.0

NumPy 1.22.0 Release Notes

NumPy 1.22.0 is a big release featuring the work of 153 contributors spread over 609 pull requests. There have been many improvements, highlights are:

  • Annotations of the main namespace are essentially complete. Upstream is a moving target, so there will likely be further improvements, but the major work is done. This is probably the most user visible enhancement in this release.
  • A preliminary version of the proposed Array-API is provided. This is a step in creating a standard collection of functions that can be used across application such as CuPy and JAX.
  • NumPy now has a DLPack backend. DLPack provides a common interchange format for array (tensor) data.
  • New methods for quantile, percentile, and related functions. The new methods provide a complete set of the methods commonly found in the literature.
  • A new configurable allocator for use by downstream projects.

These are in addition to the ongoing work to provide SIMD support for commonly used functions, improvements to F2PY, and better documentation.

The Python versions supported in this release are 3.8-3.10, Python 3.7 has been dropped. Note that 32 bit wheels are only provided for Python 3.8 and 3.9 on Windows, all other wheels are 64 bits on account of Ubuntu, Fedora, and other Linux distributions dropping 32 bit support. All 64 bit wheels are also linked with 64 bit integer OpenBLAS, which should fix the occasional problems encountered by folks using truly huge arrays.

Expired deprecations

Deprecated numeric style dtype strings have been removed

Using the strings "Bytes0", "Datetime64", "Str0", "Uint32", and "Uint64" as a dtype will now raise a TypeError.

(gh-19539)

Expired deprecations for loads, ndfromtxt, and mafromtxt in npyio

numpy.loads was deprecated in v1.15, with the recommendation that users use pickle.loads instead. ndfromtxt and mafromtxt were both deprecated in v1.17 - users should use numpy.genfromtxt instead with the appropriate value for the usemask parameter.

(gh-19615)

... (truncated)

Commits


Dependabot compatibility score

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](https://github.com/emarkou/Audio-auto-tagging/network/alerts).
Eleni Markou

Senior MLE @ FreeNow

GitHub Repository

audio-files convolutional-neural-networks music-audio-files music