3 ms·
Ah no I mean we can omit the whole "-ngl N" argument for now, as it is internally set to -1 by default in CPP code (instead of being 0 traditionally), and -1 me
by ngxson 1y ago
Ah no I mean we can omit the whole "-ngl N" argument for now, as it is internally set to -1 by default in CPP code (instead of being 0 traditionally), and -1 meaning offload everything to GPU
I have no idea how to specify custom layer specs with multi GPU, but that is interesting!
- danielhanchen 1y agoWAIT so GPU offloading is on by DEFAULT? Oh my fantastic! For now I have to "guess" via a Python script - ie I sum sum up all the .gguf split files in filesize, then detect CUDA memory usage, and specify approximately how many GPUs ie --device CUDA0,CUDA1 etc
- ngxson 1y agoAhhh no sorry I forgot that the actual code controlling this is inside llama-model.cpp ; sorry for the misinfo, the -ngl only set to max by default if you're using Metal backend (See the code in side llama_model_default_params())
- danielhanchen 1y agoOh no worries! I re-edited my comment to account for it :)