Added float16 datatype

[brownbaerchen]

Due Diligence

• General:
  • title of the PR is suitable to appear in the Release Notes
• Implementation:
  • unit tests: all split configurations tested
  • unit tests: multiple dtypes tested
  • NEW unit tests: MPS tested (1 MPI process, 1 GPU)
  • benchmarks: created for new functionality
  • benchmarks: performance improved or maintained
  • documentation updated where needed

Description

Remember the single rack exascale machine with the tiny caveat that it reaches 1.4 exaflops only in 4 bit precision? Well, for half precision, they still manage 360 petaflops, whereas for single precision they quote a lousy 5 petaflops. Maybe I am misinterpreting their specifications. But, anyways, supporting half precision, and in the future also no precision, is clearly important to stay cool.

In this PR, I simply add ht.float16 and the alias ht.half and a basic test as for the other types. I am not sure if this is enough. On my laptop all tests pass, but I have to do the PR to trigger the CI and see if there are more failures in other places.

I noticed that adding datatypes is made very difficult by somewhat hard-to-handle tables. The functionality also depends on the order of rows and columns of the tables, which is not obvious at all. I refactored the file core/types.py a lot to get rid of all this stuff.
Now, when adding a datatype, all information is handled in the datatype class itself. In particular, casting rules and mappings.

Issue/s resolved: #94

Changes proposed:

• Add ht.float16 and ht.half
• Refactor of core/types.py

Type of change

New feature (non-breaking change which adds functionality)

[github-actions]

Thank you for the PR!

[codecov]

Codecov Report

❌ Patch coverage is 95.00000% with 6 lines in your changes missing coverage. Please review.
✅ Project coverage is 91.68%. Comparing base (55d07f1) to head (414d508).
⚠️ Report is 26 commits behind head on main.

Files with missing lines	Patch %	Lines
heat/core/types.py	94.73%	6 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2092      +/-   ##
==========================================
- Coverage   91.69%   91.68%   -0.01%     
==========================================
  Files          89       89              
  Lines       13942    13999      +57     
==========================================
+ Hits        12784    12835      +51     
- Misses       1158     1164       +6     

Flag	Coverage Δ
unit	91.68% <95.00%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions]

Thank you for the PR!

[github-actions]

Thank you for the PR!

[github-actions]

Thank you for the PR!

[Copilot]

Pull request overview

This pull request adds support for the float16 (half precision) datatype to HeAT and performs a significant refactoring of the type system to make adding new datatypes easier.

Changes:

• Added ht.float16 and ht.half datatype classes with appropriate casting rules
• Refactored type management from static lookup tables to dynamic class-based attributes
• Updated type checking functions to use the new class-based system

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 7 comments.

Show a summary per file

File	Description
heat/core/types.py	Added float16 class, refactored type system from static tables to class attributes, added can_cast_to method to datatype base class
heat/core/tests/test_types.py	Added tests for float16/half datatype, corrected same_kind casting test behavior
heat/core/statistics.py	Updated to use heat_type_is_complexfloating function instead of _complexfloating tuple
heat/core/io.py	Replaced __type_codes lookup with dynamic type name indexing
heat/core/arithmetics.py	Updated to use heat_type_is_complexfloating function instead of _complexfloating tuple

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[github-actions]

Thank you for the PR!

[github-actions]

Thank you for the PR!

[github-actions]

Thank you for the PR!

[github-actions]

Thank you for the PR!