From e7e4df031b9e29d4b55a4e0b0295187f6b213db1 Mon Sep 17 00:00:00 2001 From: slaren Date: Fri, 12 Jan 2024 20:07:38 +0100 Subject: llama : ggml-backend integration (#4766) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov Co-authored-by: Johannes Gäßler --- common/common.h | 1 + 1 file changed, 1 insertion(+) (limited to 'common/common.h') diff --git a/common/common.h b/common/common.h index 1359e76a..f29be5b5 100644 --- a/common/common.h +++ b/common/common.h @@ -59,6 +59,7 @@ struct gpt_params { float p_split = 0.1f; // speculative decoding split probability int32_t n_gpu_layers = -1; // number of layers to store in VRAM (-1 - use default) int32_t n_gpu_layers_draft = -1; // number of layers to store in VRAM for the draft model (-1 - use default) + llama_split_mode split_mode = LLAMA_SPLIT_LAYER; // how to split the model across GPUs int32_t main_gpu = 0; // the GPU that is used for scratch and small tensors float tensor_split[LLAMA_MAX_DEVICES] = {0}; // how split tensors should be distributed across GPUs int32_t n_beams = 0; // if non-zero then use beam search of given width. -- cgit v1.2.3