vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Merge branch 'sycl' of https://github.com/abhilash1910/llama.cpp into sycl

Browse source
This commit is contained in:
jianyuzh 2024-01-23 14:16:24 +08:00
commit 27c08c0429
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
common/common.cpp
View file

@ -659,7 +659,7 @@ bool gpt_params_parse_ex(int argc, char ** argv, gpt_params & params) {
params.mul_mat_q = false; params.mul_mat_q = false;
#else #else
fprintf(stderr, "warning: llama.cpp was compiled without cuBLAS or SYCL. Disabling mul_mat_q kernels has no effect.\n"); fprintf(stderr, "warning: llama.cpp was compiled without cuBLAS or SYCL. Disabling mul_mat_q kernels has no effect.\n");
#endif // GGML_USE_CUBLAS #endif // GGML_USE_CUBLAS || GGML_USE_SYCL
} else if (arg == "--no-mmap") { } else if (arg == "--no-mmap") {
params.use_mmap = false; params.use_mmap = false;
} else if (arg == "--numa") { } else if (arg == "--numa") {
@ -1021,6 +1021,7 @@ void gpt_print_usage(int /*argc*/, char ** argv, const gpt_params & params) {
printf(" fraction of the model to offload to each GPU, comma-separated list of proportions, e.g. 3,1\n"); printf(" fraction of the model to offload to each GPU, comma-separated list of proportions, e.g. 3,1\n");
printf(" -mg i, --main-gpu i the GPU to use for the model (with split-mode = none),\n"); printf(" -mg i, --main-gpu i the GPU to use for the model (with split-mode = none),\n");
printf(" or for intermediate results and KV (with split-mode = row) (default: %d)\n", params.main_gpu); printf(" or for intermediate results and KV (with split-mode = row) (default: %d)\n", params.main_gpu);
#endif // LLAMA_SUPPORTS_GPU_OFFLOAD
#ifdef GGML_USE_CLBLAS #ifdef GGML_USE_CLBLAS
printf(" -nommq, --no-mul-mat-q\n"); printf(" -nommq, --no-mul-mat-q\n");
printf(" use " GGML_CUBLAS_NAME " instead of custom mul_mat_q " GGML_CUDA_NAME " kernels.\n"); printf(" use " GGML_CUBLAS_NAME " instead of custom mul_mat_q " GGML_CUDA_NAME " kernels.\n");
@ -1031,7 +1032,6 @@ void gpt_print_usage(int /*argc*/, char ** argv, const gpt_params & params) {
printf(" use " GGML_SYCL_NAME " instead of custom mul_mat_q " GGML_SYCL_NAME " kernels.\n"); printf(" use " GGML_SYCL_NAME " instead of custom mul_mat_q " GGML_SYCL_NAME " kernels.\n");
printf(" Not recommended since this is both slower and uses more VRAM.\n"); printf(" Not recommended since this is both slower and uses more VRAM.\n");
#endif // GGML_USE_SYCL #endif // GGML_USE_SYCL
#endif
printf(" --verbose-prompt print a verbose prompt before generation (default: %s)\n", params.verbose_prompt ? "true" : "false"); printf(" --verbose-prompt print a verbose prompt before generation (default: %s)\n", params.verbose_prompt ? "true" : "false");
printf(" --no-display-prompt don't print prompt at generation (default: %s)\n", !params.display_prompt ? "true" : "false"); printf(" --no-display-prompt don't print prompt at generation (default: %s)\n", !params.display_prompt ? "true" : "false");
printf(" -gan N, --grp-attn-n N\n"); printf(" -gan N, --grp-attn-n N\n");