* local metal on metal in uop syntax
* TODO: just put the axis_info in the kernelinfo
* local
* amd_matmul works @ 28 TFLOPS
* clean up matmul
* kernel8 works
* remove that
* locals
* axistype innovation
* work
* cleanup
* kernel3 regs
* cleanup kernel3
* work
* why is it broken
* no beam
* reenable
* permutes
* file path as input and have parse be in OnnxRunner.__init__
* modelproto_to_onnxrunner -> modelproto_to_runner
* whoops, fix import
* oh flakiness again, is it because it's getting gc-ed?
* small changes
* CI flaky so just move compile4 fix in
* copy typing of onnx_load
* actually can just import onnx_load instead of onnx.load
* fix external_benchmark_openpilot
* fix onnx_runner test to use onnx_helper
* rerun CI
* try run_modelproto
* spam CI a few times
* revert run_modelproto since that's flaky also
* no external onnx_load usage except onnx.py
* cursor tab complete is evil. Snuck a darn sorted in. But does order change result? Why?
* model_benchmark 193s -> 80s, add OnnxRunner.to()...
* minimize diff and clean up
* device can be None, weird but eh
---------
Co-authored-by: chenyu <chenyu@fastmail.com>
* kernel.py no longer permutes reduce axis [pr]
* delete tests that handcode uops
* regen of sops is broken...
* put import back
* just remove that
* disable those tests
* fix extract_dataset + tests
* add CI
* sops.gz itself is same as master
* yml + gzip -c + ge
* don't commit that
* bump limit to 1000
* axis=7
* test_tiny
* squash commits
* temp fix for const tensor
* actually realizing float16 can only happen in raw_data
* .float -> cast(float) to rerun CI
---------
Co-authored-by: chenyu <chenyu@fastmail.com>
* fix
* add early verbose demo test
* is this how to write tests :s
* is definition drift even a thing? gemini says it is
* clean up
* better
* even better
* try add to CI
* doesn't work quite yet
* much more work to be done
* whoops
* partition the test heh
* skipif
* some nits for better names
* add webgpu test for onnxrunner
* fix reference links
* flush for now
* start
* still need the float16 workaround in
* tiny nit for correctness
* idk hacks, I need to understand this device stuff better
* no-op?
* remove that assert for true nooooooop
* add fallback_context
* don't run linearize().uop tests in get_action_space test
this part takes 2 minutes in CI and has nothing to do with action space. also not sure if the "for some reason" comment is still relevant
* -n=auto test/models
* Implement private _linalg_eigh function for tensor eigenvalue decomposition in torch backend
* Add unit test for linalg.eigh function in TestTorchBackend
This test verifies the eigenvalue decomposition of a 2x2 tensor using the linalg.eigh function, ensuring the computed eigenvalues and reconstructed tensor match the expected results.
