tenstorrent / tt-metal

:metal: TT-NN operator library, and TT-Metalium low level kernel programming model.
Apache License 2.0
396 stars 48 forks source link

Sweeps: Default tt-smi args in CI will only reset device 0 #12006

Open jdesousa-TT opened 2 weeks ago

jdesousa-TT commented 2 weeks ago

cc: @xanderchin @tt-rkim

tt-rkim commented 2 weeks ago

You should assume the reset scripts are able to reset the machine as expected. For example, we have special reset scripts for T3000 and Galaxy. This is what conftest.py:reset_tensix does because it knows that the reset script is correct for the machine. If it's not correct, then we set it up wrong. I can explain further how exactly we know which reset script to install, if you'd like. The install method is quite automated.

Are there machines that you see that have multiple devices that aren't reset properly by the reset.sh?

cc: @TT-billteng