torch-sparse

PyPI

Is torch-sparse safe to use?

Based on the latest brin safety scan, no vulnerabilities or threats were detected for torch-sparse v0.6.18. Trust score: 95/100. No known CVE vulnerabilities, no detected threat patterns, and no suspicious capabilities identified. This is an automated, point-in-time assessment.

Install (safety-checked)

torch-sparse Passed Security Checks

No security concerns detected

clean
CVEs

0

Threats

0

Install Scripts

0

No Concerns Detected

No security concerns detected in the latest brin assessment. This is an automated, point-in-time evaluation — security posture may change.

This is an automated, point-in-time assessment and may contain errors. Findings are risk indicators, not confirmed threats. Security posture may change over time. Maintainers can dispute findings via the brin review process.

torch-sparse Capabilities & Permissions

What torch-sparse can access when installed. Review these capabilities before using with AI agents like Cursor, Claude Code, or Codex.

Filesystem Access

Reads from the filesystem.

/usr/ (r)

Environment Variables

Accesses the following environment variables.

BUILD_DOCSFORCE_CUDAFORCE_ONLY_CPUFORCE_ONLY_CUDANVCC_FLAGSWITH_METISWITH_MTMETISWITH_SYMBOLS

Native Modules

Contains native code that runs outside the JavaScript sandbox.

native extensionnumpyscipy

AGENTS.md for torch-sparse

Good instructions lead to good results. brin adds torch-sparse documentation to your AGENTS.md so your agent knows how to use it properly—improving both safety and performance.

brin init

Vercel's research: 100% accuracy with AGENTS.md vs 53% without →

torch-sparse Documentation & Source Code

For the full torch-sparse README, API documentation, and source code, visit the official package registry.

Frequently asked questions about torch-sparse safety

Weekly Downloads

8.6K

Version

0.6.18

Last Scanned

Feb 12, 2026

Trust Score

95/100·Legitimacy signals, not safety

Capabilities

Filesystem

Reads files

Environment

Accesses: BUILD_DOCS, FORCE_CUDA, FORCE_ONLY_CPU...

Native

Contains native modules