llama.cpp/ggml-cuda.h

57 lines
2.1 KiB
C
Raw Normal View History

#pragma once
#include "ggml.h"
#include "ggml-backend.h"
#ifdef GGML_USE_HIPBLAS
#define GGML_CUDA_NAME "ROCm"
#define GGML_CUBLAS_NAME "hipBLAS"
#else
#define GGML_CUDA_NAME "CUDA"
#define GGML_CUBLAS_NAME "cuBLAS"
#endif
#ifdef __cplusplus
extern "C" {
#endif
#define GGML_CUDA_MAX_DEVICES 16
// Always success. To check if CUDA is actually loaded, use `ggml_cublas_loaded`.
GGML_API void ggml_init_cublas(void);
// Returns `true` if there are available CUDA devices and cublas loads successfully; otherwise, it returns `false`.
GGML_API bool ggml_cublas_loaded(void);
GGML_API void * ggml_cuda_host_malloc(size_t size);
GGML_API void ggml_cuda_host_free(void * ptr);
GGML_API bool ggml_cuda_can_mul_mat(const struct ggml_tensor * src0, const struct ggml_tensor * src1, struct ggml_tensor * dst);
GGML_API void ggml_cuda_set_tensor_split(const float * tensor_split);
GGML_API void ggml_cuda_transform_tensor(void * data, struct ggml_tensor * tensor);
GGML_API void ggml_cuda_free_data(struct ggml_tensor * tensor);
GGML_API void ggml_cuda_assign_buffers(struct ggml_tensor * tensor);
GGML_API void ggml_cuda_assign_buffers_no_scratch(struct ggml_tensor * tensor);
GGML_API void ggml_cuda_assign_buffers_force_inplace(struct ggml_tensor * tensor);
GGML_API void ggml_cuda_assign_buffers_no_alloc(struct ggml_tensor * tensor);
GGML_API void ggml_cuda_assign_scratch_offset(struct ggml_tensor * tensor, size_t offset);
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
GGML_API void ggml_cuda_copy_to_device(struct ggml_tensor * tensor);
GGML_API void ggml_cuda_set_main_device(int main_device);
GGML_API void ggml_cuda_set_mul_mat_q(bool mul_mat_q);
GGML_API void ggml_cuda_set_scratch_size(size_t scratch_size);
GGML_API void ggml_cuda_free_scratch(void);
GGML_API bool ggml_cuda_compute_forward(struct ggml_compute_params * params, struct ggml_tensor * tensor);
GGML_API int ggml_cuda_get_device_count(void);
GGML_API void ggml_cuda_get_device_description(int device, char * description, size_t description_size);
// backend API
GGML_API ggml_backend_t ggml_backend_cuda_init(void); // TODO: take a list of devices to use
#ifdef __cplusplus
}
#endif