13 lines
337 B
C
13 lines
337 B
C
#pragma once
|
|
|
|
#include "llama-context.h"
|
|
#include "ggml.h"
|
|
#include "stdint.h"
|
|
|
|
// Reserve a new compute graph. It is valid until the next call to llama_graph_reserve.
|
|
LLAMA_API struct ggml_cgraph * llama_graph_reserve(
|
|
struct llama_context * ctx,
|
|
uint32_t n_tokens,
|
|
uint32_t n_seqs,
|
|
uint32_t n_outputs);
|