Match PyTorch arange signature

[mawad-amd]

Motivation

Match PyTorch arange signature.

Technical Details

Test Plan

Test Result

Submission Checklist

• Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.

[Copilot]

Pull Request Overview

This PR updates the Iris framework to match PyTorch's arange signature and improve memory format handling. The changes align the framework more closely with PyTorch's tensor creation APIs while maintaining compatibility with the symmetric heap allocation system.

Key changes include:

• Modified arange method signature to match PyTorch with keyword-only parameters
• Enhanced zeros_like to support additional memory formats like channels_last and channels_last_3d
• Improved device handling to default to input tensor's device rather than Iris device

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 4 comments.

File	Description
iris/iris.py	Updated arange signature, enhanced memory format support in zeros_like, improved device validation
tests/unittests/test_arange.py	New comprehensive test file for arange functionality covering all parameter combinations
tests/unittests/test_zeros_like.py	Updated tests to reflect new device defaults and added tests for new memory format support

Comments suppressed due to low confidence (1)

tests/unittests/test_zeros_like.py:1

• These tests expect ValueError to be raised, but this contradicts PyTorch's behavior. PyTorch arange(5, 1) returns an empty tensor tensor([], dtype=torch.int64) rather than raising an error. The tests should verify the actual PyTorch behavior or the implementation should be updated to match PyTorch.

# SPDX-License-Identifier: MIT