ENH: Generation of compatable device and dtype pairs for test_from_dlpack#443
Open
prady0t wants to merge 1 commit into
Open
ENH: Generation of compatable device and dtype pairs for test_from_dlpack#443prady0t wants to merge 1 commit into
test_from_dlpack#443prady0t wants to merge 1 commit into
Conversation
Signed-off-by: Pradyot Ranjan <99216956+pradyotRanjan@users.noreply.github.com>
ev-br
requested changes
Jul 24, 2026
| except Exception: | ||
| return False | ||
| else: | ||
| return True |
Member
There was a problem hiding this comment.
This actually does not work, I'm afraid: with torch, we need to filter out the device="meta", which a version on main does, cf test_dlpack.py::_compatible_devices.
With this PR (note device="meta")
In [3]: hh.device_dtype_pairs
Out[3]: sampled_from([(torch.bool, device(type='cpu')), (torch.bool, device(type='meta')), (torch.uint8, device(type='cpu')), (torch.uint8, device(type='meta')), (torch.uint16, device(type='cpu')), (torch.uint16, device(type='meta')), (torch.uint32, device(type='cpu')), (torch.uint32, device(type='meta')), (torch.uint64, device(type='cpu')), (torch.uint64, device(type='meta')), (torch.int8, device(type='cpu')), (torch.int8, device(type='meta')), (torch.int16, device(type='cpu')), (torch.int16, device(type='meta')), (torch.int32, device(type='cpu')), (torch.int32, device(type='meta')), (torch.int64, device(type='cpu')), (torch.int64, device(type='meta')), (torch.float32, device(type='cpu')), (torch.float32, device(type='meta')), (torch.float64, device(type='cpu')), (torch.float64, device(type='meta')), (torch.complex64, device(type='cpu')), (torch.complex64, device(type='meta')), (torch.complex128, device(type='cpu')), (torch.complex128, device(type='meta'))])
and
In [5]: import array_api_compat.torch as xp
In [6]: xp.from_dlpack(xp.asarray([], dtype=xp.bool), device=xp.device("meta"))
---------------------------------------------------------------------------
BufferError Traceback (most recent call last)
Cell In[6], line 1
----> 1 xp.from_dlpack(xp.asarray([], dtype=xp.bool), device=xp.device("meta"))
File ~/miniforge3/envs/array-api/lib/python3.12/site-packages/torch/utils/dlpack.py:149, in from_dlpack(ext_tensor, device, copy)
146 raise AssertionError(f"from_dlpack: unsupported device type: {type(device)}")
148 # Convert target device to DLPack format
--> 149 target_dl_device = torch._C._torchDeviceToDLDevice(device)
151 # Detect cross-device transfer by comparing source and target devices
152 # E.g. CPU->CUDA, cuda:0->cuda:1, etc.
153 cross_device_transfer = (ext_device != target_dl_device)
BufferError: Cannot pack tensors on meta
which leads to failing
$ ARRAY_API_TESTS_MODULE=array_api_compat.torch pytest array_api_tests/test_dlpack.py::test_from_dlpack -v --max-examples=10_000 -vs
|
|
||
| _dtype_categories = [(xp.bool,), dh.uint_dtypes, dh.int_dtypes, dh.real_float_dtypes, dh.complex_dtypes] | ||
| _sorted_dtypes = [d for category in _dtype_categories for d in category] | ||
| _device_dtype_pairs = [(d, device) for d in _sorted_dtypes for device in xp.__array_namespace_info__().devices() if dh.is_dtype_device_compatible(d, device)] |
Member
There was a problem hiding this comment.
since this PR needs a tweak anyway, please break the long long line
|
|
||
| @given( | ||
| x=hh.arrays(dtype=hh.all_dtypes, shape=hh.shapes(min_dims=1, max_side=2)), | ||
| dtype_device_pair = hh.device_dtype_pairs, |
Member
There was a problem hiding this comment.
I'm a kind of sad to not being able to use x = hh.arrays(....) in the decorator, but well, "what is, is".
Would be nice to come up with a not-very-ugly general pattern for #302, but taht's not a blocker here. At this stage, let's see what it takes to fix an immediate problem, and keep thinking about a general solution for a general problem on a back-burner.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #440
This is a fairly minimal fix. The new
hh.device_dtype_pairsstrategy generates pairs of all possible (dtype, device) such that dtype is compatible with device.Running:
succeeds, which previously failed for
dtype = float64ondevice = no_float64