macOS build #7
macOS build #7
Conversation
ericcurtin
force-pushed
the
mac
branch
8 times, most recently
from
17e3b38
to
473ed5f
Compare
ericcurtin
force-pushed
the
mac
branch
6 times, most recently
from
784b8dd
to
9a94daf
Compare
install.sh
Outdated
| elif available nvidia-smi; then | ||
| nvidia_available="true" | ||
| fi | ||
| # gpu_check() { |
There was a problem hiding this comment.
Why are you commenting this out?
There was a problem hiding this comment.
I never completed the GPU checking stuff... Trying to get the design correct first before bringing on the maintenance of all the GPU backends, testing matrix can get big fast.
And because I started running shellcheck on everything, shellcheck was identifying this stuff as unused shell scripts (correctly)... Commenting this out makes the builds green
Maybe I should just delete this stuff...
There was a problem hiding this comment.
I am fine with commenting it out.
install.sh
Outdated
| # gpu_check() { | ||
| # if available lspci && lspci -d '10de:' | grep -q 'NVIDIA'; then | ||
| # nvidia_available="true" | ||
| # elif available lshw && nvidia_lshw; then |
There was a problem hiding this comment.
This line looks wrong.
Shouldn't this be
if available nvidia_lshw && nvidia_lshw; then?
There was a problem hiding this comment.
nvidia_lshw is a bash function a couple of lines up
lshw is a binary required to query for GPU type (there's other ways of querying GPU type, it's one of the ways of doing it rootless), not always installed.
We know nvidia_lshw is available, it's declared in this function.
install.sh
Outdated
| } | ||
| # if available lspci && lspci -d '1002:' | grep -q 'AMD'; then | ||
| # amd_available="true" | ||
| # elif available lshw && amd_lshw; then |
There was a problem hiding this comment.
Ditto
if available amd_lshw && amd_lshw; then?
To ensure things work on macOS. Signed-off-by: Eric Curtin <[email protected]>
|
Deleting commented out lines for now. When we get around to enhancing this more below are some decent reference scripts, we may have to write in a portable POSIX-like manner to be compatible with bash for Linux and zsh for macOS. Eventually we will have to rpm package properly, but it's really not straightforward for these AI/LLM tools, optional dependencies based on hardware, instable APIs that need to be changed regularly. The AI world is kinda young also, so we really want to be up-to date with things like llama.cpp upstream to be compatible with the latest LLMs like Granite Code, Llama 3 as examples. https://github.com/mudler/LocalAI/blob/master/docs/static/install.sh |
|
LGTM |
To ensure things work on macOS.