Stabilize tensor-train relative truncation

[FlorianPfaff]

Bug

TensorTrain.from_dense(...) computed the Frobenius norm with an unscaled Euclidean norm, and _choose_rank(...) squared raw singular values while evaluating truncation tails.

For large but finite tensors, those intermediate operations overflow even when the tensor, requested relative tolerance, and correct tensor-train decomposition are finite. For example, diag(1e200, 1e200) raises under strict NumPy overflow handling; without strict handling, the resulting infinite tolerance can incorrectly collapse the decomposition to rank one.

Fix

• compute vector norms after scaling by the largest magnitude;
• compute singular-value tail energies in normalized coordinates;
• compare against the correspondingly scaled local tolerance;
• preserve zero-tolerance and max_rank behavior.

Regression coverage

A focused test constructs diag(1e200, 1e200) under np.errstate(over="raise", invalid="raise"). It verifies ranks (1, 2, 1) and reconstruction of the finite input.

Validation

• refreshed directly onto current main (f196ec41d1d8681e279936db9f1a0a3abfa276e7);
• final commit changes only the tensor-train helper and its focused regression test;
• supersedes Stabilize tensor-train relative truncation #5249, whose test workflow ran against an older base with a now-fixed APT setup;
• GitHub Actions is authoritative for the full repository matrix.

[github-actions]

✅MegaLinter analysis: Success

Descriptor	Linter	Files	Fixed	Errors	Warnings	Elapsed time
✅ COPYPASTE	jscpd	yes		no	no	27.83s
✅ JSON	prettier	7	0	0	0	1.08s
✅ JSON	v8r	7		0	0	4.05s
✅ MARKDOWN	markdownlint	68	0	0	0	2.0s
✅ MARKDOWN	markdown-table-formatter	68	0	0	0	0.8s
✅ PYTHON	black	1866	87	0	0	98.53s
✅ PYTHON	isort	1866	140	0	0	2.89s
✅ REPOSITORY	betterleaks	yes		no	no	2.46s
✅ REPOSITORY	checkov	yes		no	no	55.23s
✅ REPOSITORY	gitleaks	yes		no	no	17.91s
✅ REPOSITORY	git_diff	yes		no	no	0.33s
✅ REPOSITORY	secretlint	yes		no	no	84.1s
✅ REPOSITORY	syft	yes		no	no	6.0s
✅ REPOSITORY	trivy-sbom	yes		no	no	6.19s
✅ REPOSITORY	trufflehog	yes		no	no	32.61s
✅ YAML	prettier	11	0	0	0	0.79s
✅ YAML	v8r	11		0	0	11.59s
✅ YAML	yamllint	11		0	0	0.62s

Notices

📣 MegaLinter 9.5.0 is out! Discover the new features and security recommendations in the release announcement. (Skip this info by defining SECURITY_SUGGESTIONS: false)

See detailed reports in MegaLinter artifacts

Your project could benefit from a custom flavor, which would allow you to run only the linters you need, and thus improve runtime performances. (Skip this info by defining FLAVOR_SUGGESTIONS: false)

• Documentation: Custom Flavors
• Command: npx mega-linter-runner@9.6.0 --custom-flavor-setup --custom-flavor-linters PYTHON_BLACK,PYTHON_ISORT,COPYPASTE_JSCPD,JSON_V8R,JSON_PRETTIER,MARKDOWN_MARKDOWNLINT,MARKDOWN_MARKDOWN_TABLE_FORMATTER,REPOSITORY_CHECKOV,REPOSITORY_GIT_DIFF,REPOSITORY_GITLEAKS,REPOSITORY_BETTERLEAKS,REPOSITORY_SECRETLINT,REPOSITORY_SYFT,REPOSITORY_TRIVY_SBOM,REPOSITORY_TRUFFLEHOG,YAML_PRETTIER,YAML_YAMLLINT,YAML_V8R

Show us your support by starring ⭐ the repository