Commit History

Autor SHA1 Mensaxe Data
  Daniel Hiltgen 9754c6d9d8 Harden AMD driver lookup logic hai 1 ano
  Daniel Hiltgen 6d84f07505 Detect AMD GPU info via sysfs and block old cards hai 1 ano
  Daniel Hiltgen 4072b5879b Merge pull request #2246 from dhiltgen/reject_cuda_without_avx hai 1 ano
  Daniel Hiltgen 15562e887d Don't disable GPUs on arm without AVX hai 1 ano
  Daniel Hiltgen f07f8b7a9e Harden for zero detected GPUs hai 1 ano
  Daniel Hiltgen e02ecfb6c8 Merge pull request #2116 from dhiltgen/cc_50_80 hai 1 ano
  Daniel Hiltgen 667a2ba18a Detect lack of AVX and fallback to CPU mode hai 1 ano
  Daniel Hiltgen 9d7b5d6c91 Ignore AMD integrated GPUs hai 1 ano
  Daniel Hiltgen 013fd07139 More logging for gpu management hai 1 ano
  Daniel Hiltgen 987c16b2f7 Report more information about GPUs in verbose mode hai 1 ano
  Daniel Hiltgen a447a083f2 Add compute capability 5.0, 7.5, and 8.0 hai 1 ano
  Jeffrey Morgan f32ea81b21 increase minimum overhead to 1024MiB (#2114) hai 1 ano
  Daniel Hiltgen 681a914990 Add support for CUDA 5.2 cards hai 1 ano
  Daniel Hiltgen 552db98bf1 More WSL paths hai 1 ano
  Self Denial eb76f3e379 Fix CPU-only build under Android Termux enviornment. hai 1 ano
  Daniel Hiltgen abec7f06e5 Merge pull request #2056 from dhiltgen/slog hai 1 ano
  Daniel Hiltgen fedd705aea Mechanical switch from log to slog hai 1 ano
  Alexander F. Rødseth f4bf1d514f Let gpu.go and gen_linux.sh also find CUDA on Arch Linux hai 1 ano
  Daniel Hiltgen d88c527be3 Build multiple CPU variants and pick the best hai 1 ano
  Daniel Hiltgen 8da7bef05f Support multiple variants for a given llm lib type hai 1 ano
  Jeffrey Morgan b24e8d17b2 Increase minimum CUDA memory allocation overhead and fix minimum overhead for multi-gpu (#1896) hai 1 ano
  Daniel Hiltgen 3c49c3ab0d Harden GPU mgmt library lookup hai 1 ano
  Jeffrey Morgan c336693f07 calculate overhead based number of gpu devices (#1875) hai 1 ano
  Daniel Hiltgen 1961a81f03 Set corret CUDA minimum compute capability version hai 1 ano
  Jeffrey Morgan 6df83e6daa update rough cuda overhead estimate to 15% + 384MiB hai 1 ano
  Jeffrey Morgan 6164f378f2 revert cuda overhead to 20% hai 1 ano
  Jeffrey Morgan 6566387ae3 add `TODO` for cuda overhead hai 1 ano
  Jeffrey Morgan 37708931fb update cuda overhead to 20% to fix crashes when switching between models and large context sizes hai 1 ano
  Jeffrey Morgan f6cb0a553c update cuda overhead to 15% or 400MiB hai 1 ano
  Jeffrey Morgan 2680078c13 fix build on linux hai 1 ano