ȅiS%SSKJr SSKJrJr SSKrSSKJr /SQr S\ 4Sjr \S\ 4Sj5r \S \S \S\ 4S j5r\SS \S\ 4S jj5r\S\4S j5r\S\4Sj5rSq\\ \S'SSjrg)) lru_cache)Optional TYPE_CHECKINGN)Library)get_core_countget_nameis_built is_availableis_macos13_or_neweris_macos_or_newerreturnc6[RR$)zReturn whether PyTorch is built with MPS support. Note that this doesn't necessarily mean MPS is available; just that if this PyTorch binary were run a machine with working MPS drivers and devices, we would be able to use it. )torch_C_has_mpsU/home/james-whalen/.local/lib/python3.13/site-packages/torch/backends/mps/__init__.pyr r s 88  rc>[RR5$)z7Return a bool indicating if MPS is currently available.)rr_mps_is_availablerrrr r s 88 % % ''rmajorminorc@[RRX5$)zHReturn a bool indicating whether MPS is running on given MacOS or newer.rr_mps_is_on_macos_or_newer)rrs rr r "s 88 - -e ;;rcB[RRSU5$)zEReturn a bool indicating whether MPS is running on MacOS 13 or newer. r)rs rr r (s 88 - -b% 88rc>[RR5$)zReturn Metal device name)rr _mps_get_namerrrrr.s 88 ! ! ##rc>[RR5$)zReturn GPU core count. According to the documentation, one core is comprised of 16 Execution Units. One execution Unit has 8 ALUs. And one ALU can run 24 threads, i.e. one core is capable of executing 3072 threads concurrently. )rr_mps_get_core_countrrrrr4s 88 ' ' ))r_libc[c[5(dgSSKJn SSKJn [ SS5q[RSUS5 [RS US5 g) zr7s-* - $ (d( (