entmax

library1.3pypypiunverified

A family of sparse alternatives to softmax, including entmax, sparsemax, normmax, and budget, with corresponding loss functions. Current version: 1.3. Release cadence is irregular, with updates as needed for bug fixes and new activations.

ai-ml

pip install entmax

github ↗package ↗

INSTALL

IMPORT

SIG · ENTMAX

entmax

ai-mlpythonv1.3

Install

67.3s avg

Import

—

Disk

4787MB

Pass rate

4/ 10

Env Coverage4 / 10

glibc

3.9–3.13

musl

3.9–3.13

Install & Compatibility

Where this runs

tested against v1.3 · pip install

no network on importno background threads

Install × environment matrix

Each cell = how many times install + import succeeded across repeated harness runs. Partial = flaky.
glibc = Debian/Ubuntu slim · musl = Alpine Linux

musl

glibc

py 3.10

✕ build_error

✓ 74.5s

py 3.11

✕ build_error

✓ 69.5s

py 3.12

✕ build_error

✓ 67.3s

py 3.13

✕ build_error

✓ 57.8s

py 3.9

✕ build_error

✕ timeout

4787MB installed

● package 4787MB

Code

Verified usage

Verified import paths — ran on the pinned version, not inferred.

entmax_loss

✓ from entmax import entmax_loss

Official import path for the entmax loss function.

Sparsemax

✓ from entmax import Sparsemax

Official import path for the Sparsemax activation layer.

Apply entmax-1.5 activation and the generic bisection-based entmax function.

import torch from entmax import entmax15, entmax_bisect x = torch.randn(2, 5) y = entmax15(x, dim=-1) print(y) z = entmax_bisect(x, alpha=1.5, dim=-1) print(z)

Debug

Known issues

gotchaentmax 1.x requires PyTorch to be installed before installation. Attempting to install entmax without torch will fail. Use conda or pip to install PyTorch first.

fix

Install PyTorch first: pip install torch. Then proceed with pip install entmax.

affects: <=1.3

gotchaThe bisection-based functions (entmax_bisect, etc.) are slower than the closed-form variants. They are intended for non-standard alpha values (not 1.5 or 2).

fix

Use entmax15 or Sparsemax for standard alphas; use bisection only for custom alphas.

affects: >=1.0

deprecatedThe 'alpha' parameter in some older functions is deprecated; use the specific entmax15 or Sparsemax functions instead.

fix

Migrate to entmax15 (alpha=1.5) or Sparsemax (alpha=2).

affects: <=1.2

Upgrade

Version history

1.3latest on PyPI · released Feb 7, 2024

Audit

Dependencies

No dependency data recorded yet.

Agent activity

17 hits · last 30 days

node

OpenAI (training)

Resources

githubhttps://github.com/deep-spin/entmax ↗

packagehttps://pypi.org/project/entmax/ ↗