Guide for operator level performance debugging

1. Install pytorch with torch-xpu-ops.
2. Install torchvision/audio

pip install torchvision torchaudio --no-deps --index-url https://download.pytorch.org/whl/nightly/xpu

3. Install triton using the following command:

pip install --force-reinstall  "git+https://github.com/intel/intel-xpu-backend-for-triton@0bcc8265e677e5321606a3311bf71470f14456a8#subdirectory=python"

4. Build install torch-benchmark from source:

rm -rf benchmark && git clone https://github.com/pytorch/benchmark
cd benchmark && git checkout 373ffb19dc470f4423a3176a4133f8f4b3cdb5bd && pip install -r requirements.txt
python install.py --continue_on_fail
pip install -e .
cd ..