Setepenre
Setepenre
This is what I got on my side overall with `PYTORCH_TEST_WITH_ROCM=1 python test/run_test.py ` - test_autograd 919 tests in 161s (6 skipped) - test_cuda 154 tests in 19sec (77 skipped,...
ninja is just the build system that pytorch can use to compile itself. You do not have to use it. it is explained [here][1]. ROCm has rocblas and miopen for...
thanks, I updated the list of dependencies
nice, I updated it
Full log off an installation [pytorch_cluster.txt](https://github.com/user-attachments/files/17330796/pytorch_cluster.txt)
I tried on * 8xA100 and 8xH100, * model is `torchtune.models.llama3_1.llama3_1_70b` * dataset is torchtune.datasets.alpaca_dataset
I downgraded to `0.2.1` while I wait for pytorch to release a new version, in my case I cannot use nightlies. But if it OOM why did I not see...
https://docs.pytest.org/en/7.1.x/explanation/flaky.html#xfail-strict
``` INTERNALERROR> Traceback (most recent call last): INTERNALERROR> File "/home/newton/work/orion/.tox/py-all/lib/python3.9/site-packages/_pytest/main.py", line 270, in wrap_session INTERNALERROR> session.exitstatus = doit(config, session) or 0 INTERNALERROR> File "/home/newton/work/orion/.tox/py-all/lib/python3.9/site-packages/_pytest/main.py", line 324, in _main INTERNALERROR> config.hook.pytest_runtestloop(session=session)...
Looks to be twice as fast to insignificant