Move hf_MPT_7b_instruct to canary_models (#1810)
Summary:
hf_MPT_7b_instruct model doesn't pass CI test because of CUDA OOM on A10G: https://github.com/pytorch/benchmark/actions/runs/5764053023/job/15627183501
It will also cause unknown runner shutdown on A100: https://github.com/pytorch/benchmark/actions/runs/5764053022
Therefore, we have to move it to canary_models set.
Pull Request resolved: https://github.com/pytorch/benchmark/pull/1810
Reviewed By: msaroufim
Differential Revision: D48122323
Pulled By: xuzhao9
fbshipit-source-id: 342c9df456edbe5d440c09e041d0ad3e723a07ad
