 Daniel Hiltgen
|
6c5ccb11f9
Revamp ROCm support
|
1 rok temu |
|
Daniel Hiltgen
|
be330174dd
Allow setting max vram for workarounds
|
1 rok temu |
|
Daniel Hiltgen
|
9754c6d9d8
Harden AMD driver lookup logic
|
1 rok temu |
|
Daniel Hiltgen
|
6d84f07505
Detect AMD GPU info via sysfs and block old cards
|
1 rok temu |
 Daniel Hiltgen
|
4072b5879b
Merge pull request #2246 from dhiltgen/reject_cuda_without_avx
|
1 rok temu |
|
Daniel Hiltgen
|
15562e887d
Don't disable GPUs on arm without AVX
|
1 rok temu |
|
Daniel Hiltgen
|
f07f8b7a9e
Harden for zero detected GPUs
|
1 rok temu |
|
Daniel Hiltgen
|
e02ecfb6c8
Merge pull request #2116 from dhiltgen/cc_50_80
|
1 rok temu |
|
Daniel Hiltgen
|
667a2ba18a
Detect lack of AVX and fallback to CPU mode
|
1 rok temu |
|
Daniel Hiltgen
|
9d7b5d6c91
Ignore AMD integrated GPUs
|
1 rok temu |
|
Daniel Hiltgen
|
013fd07139
More logging for gpu management
|
1 rok temu |
|
Daniel Hiltgen
|
987c16b2f7
Report more information about GPUs in verbose mode
|
1 rok temu |
|
Daniel Hiltgen
|
a447a083f2
Add compute capability 5.0, 7.5, and 8.0
|
1 rok temu |
 Jeffrey Morgan
|
f32ea81b21
increase minimum overhead to 1024MiB (#2114)
|
1 rok temu |
|
Daniel Hiltgen
|
681a914990
Add support for CUDA 5.2 cards
|
1 rok temu |
|
Daniel Hiltgen
|
552db98bf1
More WSL paths
|
1 rok temu |
 Self Denial
|
eb76f3e379
Fix CPU-only build under Android Termux enviornment.
|
1 rok temu |
|
Daniel Hiltgen
|
abec7f06e5
Merge pull request #2056 from dhiltgen/slog
|
1 rok temu |
|
Daniel Hiltgen
|
fedd705aea
Mechanical switch from log to slog
|
1 rok temu |
 Alexander F. Rødseth
|
f4bf1d514f
Let gpu.go and gen_linux.sh also find CUDA on Arch Linux
|
1 rok temu |
|
Daniel Hiltgen
|
d88c527be3
Build multiple CPU variants and pick the best
|
1 rok temu |
|
Daniel Hiltgen
|
8da7bef05f
Support multiple variants for a given llm lib type
|
1 rok temu |
|
Jeffrey Morgan
|
b24e8d17b2
Increase minimum CUDA memory allocation overhead and fix minimum overhead for multi-gpu (#1896)
|
1 rok temu |
|
Daniel Hiltgen
|
3c49c3ab0d
Harden GPU mgmt library lookup
|
1 rok temu |
|
Jeffrey Morgan
|
c336693f07
calculate overhead based number of gpu devices (#1875)
|
1 rok temu |
 Daniel Hiltgen
|
1961a81f03
Set corret CUDA minimum compute capability version
|
1 rok temu |
|
Jeffrey Morgan
|
6df83e6daa
update rough cuda overhead estimate to 15% + 384MiB
|
1 rok temu |
|
Jeffrey Morgan
|
6164f378f2
revert cuda overhead to 20%
|
1 rok temu |
|
Jeffrey Morgan
|
6566387ae3
add `TODO` for cuda overhead
|
1 rok temu |
|
Jeffrey Morgan
|
37708931fb
update cuda overhead to 20% to fix crashes when switching between models and large context sizes
|
1 rok temu |
