# Python/C++ API rules for device-generic APIs **URL:** \ **Category:** frontend API\ **Created:** [October 3, 2024, 3:06pm UTC](https://dev-discuss.pytorch.org/t/python-c-api-rules-for-device-generic-apis/2511 "2024-10-03T15:06:40Z")\ **Posts on this page:** 3\ **Page:** 1
**Author:** ![albanD](https://yyz2.discourse-cdn.com/flex036/user_avatar/dev-discuss.pytorch.org/alband/32/9_2.png) [@albanD](https://dev-discuss.pytorch.org/u/albanD)\ **Post date:** [October 3, 2024, 3:06pm UTC](https://dev-discuss.pytorch.org/t/python-c-api-rules-for-device-generic-apis/2511/1 "2024-10-03T15:06:40Z")
We want to follow the following rule for any new API being added (and are going to migrate existing API to follow this as well): 1. Any API that related to device property should be done as a Device-agnostic API (all devices if possible, else accelerator) unless there is a clear reason on why it does not apply to other devices. 2. Device-agnostic APIs that apply to all devices should be placed under the `torch` namespace and accept a `device_type` optional argument (with default value being the current accelerator or cpu). 3. Device-agnostic APIs for accelerators style device ([torch — PyTorch main documentation](https://pytorch.org/docs/main/torch.html#accelerators)) should be placed under the `torch.acc` namespace and not accept a `device_type` argument (they are always applied on the current accelerator). 4. Device-specific APIs should be placed under device-specific submodules (torch.cuda, torch.xpu, torch.mps, etc). Please comment here if there is any concern/feedback about this new rule. For more details on the discussions that lead to this proposal, see [[RFC] A device-agnostic Python runtime API design for stream-based accelerators · Issue #128403 · pytorch/pytorch · GitHub](https://github.com/pytorch/pytorch/issues/128403) , partial list of related issues in both core ([[RFC] A device-agnostic Python device memory related API design for stream-based accelerators · Issue #134978 · pytorch/pytorch · GitHub](https://github.com/pytorch/pytorch/issues/134978), etc) and ao ([Enable Intel GPU by dbyoung18 · Pull Request #753 · pytorch/ao · GitHub](https://github.com/pytorch/ao/pull/753)) as well as early investigative implementation ([Refine the logic of device construction when only device index is given by guangyey · Pull Request #129119 · pytorch/pytorch · GitHub](https://github.com/pytorch/pytorch/pull/129119), Stream/Event API, [Update pin memory related APIs to not pass 'device' argument by wizzniu · Pull Request #131858 · pytorch/pytorch · GitHub](https://github.com/pytorch/pytorch/pull/131858), etc). The credit for driving this discussion goes to @guangyey and other people involved on the issues listed above. ---
**Author:** ![EikanWang](https://yyz2.discourse-cdn.com/flex036/user_avatar/dev-discuss.pytorch.org/eikanwang/32/992_2.png) [@EikanWang](https://dev-discuss.pytorch.org/u/EikanWang)\ **Post date:** [October 8, 2024, 1:18pm UTC](https://dev-discuss.pytorch.org/t/python-c-api-rules-for-device-generic-apis/2511/2 "2024-10-08T13:18:30Z")
Thanks for the rules to ensure we are on the same page. It will help us move fast to implement more device-agnostic API PRs ✌ ---
**Author:** ![shink](https://yyz2.discourse-cdn.com/flex036/user_avatar/dev-discuss.pytorch.org/shink/32/1970_2.png) [@shink](https://dev-discuss.pytorch.org/u/shink)\ **Post date:** [January 23, 2025, 8:51am UTC](https://dev-discuss.pytorch.org/t/python-c-api-rules-for-device-generic-apis/2511/3 "2025-01-23T08:51:09Z")
Hey thanks for the rules! I noticed these PRs are almost landed so is there still some work to do? If so, I’d love to do it!