Enable 08-grouped-gemm on A770

[pbchekin]

This tutorial works with the current PyTorch commit id.

This reverts #2782, fixes #2737.

[pbchekin]

Workflow on A770: https://github.com/intel/intel-xpu-backend-for-triton/actions/runs/12054461624

[anmyachev]

Workflow on A770: https://github.com/intel/intel-xpu-backend-for-triton/actions/runs/12054461624

It seems that due to other errors, CI did not reach the tutorial.

[pbchekin]

Workflow on A770: https://github.com/intel/intel-xpu-backend-for-triton/actions/runs/12054461624

It seems that due to other errors, CI did not reach the tutorial.

Yep, failed on core test. I re-ran with "ignore errors" (this is how we run tests daily for A770) and 08-grouped-gemm executed this time (did not hang):
https://github.com/intel/intel-xpu-backend-for-triton/actions/runs/12054913625

tutorials: passed: 6, failed: 0, skipped: 4, xfailed: 0, total: 10, fixme: 0, pass rate (w/o xfailed): 60.0%
all: passed: 17265, failed: 2, skipped: 1641, xfailed: 2176, total: 21084, fixme: 8, pass rate (w/o xfailed): 91.31%