extern/whisper.cpp
1
0
Fork 1
mirror of https://github.com/ggerganov/whisper.cpp.git synced 2025-08-13 07:48:08 +02:00
Code Issues Packages Projects Releases Wiki Activity

talk-llama : sync llama.cpp

Browse Source
This commit is contained in:
Georgi Gerganov
2025-07-01 12:21:09 +03:00
parent c4ea72be9a
commit 1f816de7da
24 changed files with 1456 additions and 430 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
examples/talk-llama/llama-context.h
View File

@ -18,7 +18,7 @@ class llama_io_read_i;
class llama_io_write_i;
struct llama_memory_i;
struct llama_memory_state_i;
struct llama_memory_context_i;
struct llama_context {
// init scheduler and compute buffers, reserve worst-case graphs
@ -93,14 +93,14 @@ struct llama_context {
int32_t il_end);
// process a single ubatch with a specific graph type
// if memory_state is provided, it will be applied first to the context's memory
// if memory_context is provided, it will be applied first to the context's memory
// ret contains the status of the graph computation
// returns nullptr only if ret != GGML_STATUS_SUCCESS
llm_graph_result_ptr process_ubatch(
const llama_ubatch & ubatch,
llm_graph_type gtype,
llama_memory_state_i * mstate,
ggml_status & ret);
const llama_ubatch & ubatch,
llm_graph_type gtype,
llama_memory_context_i * mctx,
ggml_status & ret);
int encode(const llama_batch & batch_inp);
int decode(const llama_batch & batch_inp);
@ -197,15 +197,15 @@ public:
ggml_status graph_compute(ggml_cgraph * gf, bool batched);
// reserve a graph with a dummy ubatch of the specified size
ggml_cgraph * graph_reserve(uint32_t n_tokens, uint32_t n_seqs, uint32_t n_outputs, const llama_memory_state_i * mstate);
ggml_cgraph * graph_reserve(uint32_t n_tokens, uint32_t n_seqs, uint32_t n_outputs, const llama_memory_context_i * mctx);
private:
llm_graph_result_ptr graph_build(
ggml_context * ctx,
ggml_cgraph * gf,
const llama_ubatch & ubatch,
llm_graph_type gtype,
const llama_memory_state_i * mstate);
ggml_context * ctx,
ggml_cgraph * gf,
const llama_ubatch & ubatch,
llm_graph_type gtype,
const llama_memory_context_i * mctx);
llm_graph_cb graph_get_cb() const;