deps(whisper.cpp): update, fix cublas build

[mudler]

Description

This PR fixes #1733

Notes for Reviewers

See also: https://gitlab.kitware.com/cmake/cmake/-/issues/25536 - looks a cmake issue (?)
upstream: ggerganov/whisper.cpp#1553

PR (whisper.cpp): ggerganov/whisper.cpp#1973

Signed commits

• Yes, I signed my commits.

[netlify]

✅ Deploy Preview for localai canceled.

Name	Link
🔨 Latest commit	9ae3e67
🔍 Latest deploy log	https://app.netlify.com/sites/localai/deploys/65f82d5a440bde00086dd17f

[mudler]

opening a PR upstream with the fix too: ggerganov/whisper.cpp#1973

for reference, this is solved by adding -L$(CUDA_PATH)/stubs -lcuda to the whisper LDFLAGS and in CGO_LDFLAGS when building the golang binary with libwhisper.a (not the .so).

I've traced it back by trying to find the symbols that it complains about:

nm -D /usr/local/cuda/targets/x86_64-linux/lib/*.so   | grep cuMem
<no results>

And instead present in the stubs directory:

root@76d08ab315dc:/build# nm -D /usr/local/cuda/targets/x86_64-linux/lib/stubs/*.so   | grep cuMem                                                                                                                                           
0000000000008000 T cuMemAddressFree         
0000000000007ff0 T cuMemAddressReserve    
00000000000081d0 T cuMemAdvise                                                                                        
0000000000008f30 T cuMemAlloc
0000000000009710 T cuMemAllocAsync
00000000000080e0 T cuMemAllocAsync_ptsz
0000000000009720 T cuMemAllocFromPoolAsync
0000000000008160 T cuMemAllocFromPoolAsync_ptsz
0000000000008f70 T cuMemAllocHost
0000000000007bf0 T cuMemAllocHost_v2
0000000000007c40 T cuMemAllocManaged
0000000000008f40 T cuMemAllocPitch
... (and many others)

See also: NVIDIA/nvidia-docker#508