If you are into running local LLM models you have came across llama.cpp – basically your best friend in optimizing a model to run locally on your pc. In our case we have a laptop which runs with ~= 16 GB RAM, integrated GPU AMD Ryzen 7000 serious.
There is a GPU built into the computer, but in any case very tightly coupled with the CPU so they form one singular unit. CPU + iGPU share the same memory that we have.
This is what we initially have
free -h total used free shared buff/cache availableMem: 13Gi 4,9Gi 4,1Gi 232Mi 5,1Gi 8,5GiSwap: 511Mi 0B 511Mi
How llama.cpp help us in our case ?
The main goal of llama.cpp is to enable LLM inference with minimal setup and state-of-the-art performance on a wide range of hardware – locally and in the cloud.
- LLM inference – model being asked and generating a token.
- AVX, AVX2, AVX512 and AMX support for x86 architectures – this is like additional power for your CPU. Think about it as a way to actually handle multiple numbers at once vs a single number, when we are doing computations. In our case we have avx2.
lscpuArchitecture: x86_64 CPU op-mode(s): 32-bit, 64-bit Address sizes: 48 bits physical, 48 bits virtual Byte Order: Little EndianCPU(s): 8 On-line CPU(s) list: 0-7Vendor ID: AuthenticAMD Model name: AMD Ryzen 3 7320U with Radeon Graphics CPU family: 23 Model: 160 Thread(s) per core: 2 Core(s) per socket: 4 Socket(s): 1 Stepping: 0 Frequency boost: enabled CPU(s) scaling MHz: 35% CPU max MHz: 4151,7300 CPU min MHz: 425,1780 BogoMIPS: 4791,47 Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid extd_a picid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs s kinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb cat_l3 cdp_l3 hw_pstate ssbd mba ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr rdpru wbnoinvd cppc arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pause filter pfthreshold avic v_vmsave_vmload vgif v_spec_ctrl umip rdpid overflow_recov succor smca
- 1.5-bit, 2-bit, 3-bit, 4-bit, 5-bit, 6-bit, and 8-bit integer quantization for faster inference and reduced memory use – we are going to reduce the precision of the model a little bit, so that we could run it locally.
- llama-server – A lightweight, OpenAI API compatible, HTTP server for serving LLMs.
All right, let’s get back to work.
- Clone the repo ( this might take some time ).
git clone https://github.com/ggml-org/llama.cpp.gitCloning into 'llama.cpp'...remote: Enumerating objects: 85303, done.remote: Counting objects: 100% (74/74), done.remote: Compressing objects: 100% (48/48), done.remote: Total 85303 (delta 41), reused 26 (delta 26), pack-reused 85229 (from 2)Receiving objects: 100% (85303/85303), 338.96 MiB | 1.74 MiB/s, done.Resolving deltas: 100% (61591/61591), done.
2. Build the llama.cpp ( in our case for CPU backend, in your case it might a different backend ), also make sure you have cmake installed.
https://github.com/ggml-org/llama.cpp/blob/master/docs/build.md
Initially I had this error, but after install build-essential it was fixed.
cmake -B build -DCMAKE_BUILD_TYPE=Debug-- The C compiler identification is GNU 15.2.0-- The CXX compiler identification is unknown-- Detecting C compiler ABI info-- Detecting C compiler ABI info - done-- Check for working C compiler: /usr/bin/cc - skipped-- Detecting C compile features-- Detecting C compile features - doneCMake Error at CMakeLists.txt:2 (project): No CMAKE_CXX_COMPILER could be found. Tell CMake where to find the compiler by setting either the environment variable "CXX" or the CMake cache entry CMAKE_CXX_COMPILER to the full path to the compiler, or to the compiler name if it is in the PATH.
srejournals 
Leave a comment