alihan/whisper.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/whisper.cpp.git synced 2023-11-04 02:52:44 +03:00
Code Issues Packages Projects Releases Wiki Activity

talk-llama : add new example + sync ggml from llama.cpp (#664)

Browse Source 
* talk-llama : talk with LLaMA AI

* talk.llama : disable EOS token

* talk-llama : add README instructions

* ggml : fix build in debug
This commit is contained in:
Georgi Gerganov
2023-03-27 21:00:32 +03:00
committed by GitHub
parent 8e361d90d7
commit 4a0deb8b1e
14 changed files with 5061 additions and 528 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
ggml.h
View File

@@ -198,6 +198,8 @@ struct ggml_object;
struct ggml_context;
enum ggml_type {
GGML_TYPE_Q4_0,
GGML_TYPE_Q4_1,
GGML_TYPE_I8,
GGML_TYPE_I16,
GGML_TYPE_I32,
@@ -226,7 +228,9 @@ enum ggml_op {
GGML_OP_STEP,
GGML_OP_RELU,
GGML_OP_GELU,
GGML_OP_SILU,
GGML_OP_NORM, // normalize
GGML_OP_RMS_NORM,
GGML_OP_MUL_MAT,
@@ -326,7 +330,10 @@ void ggml_print_objects(const struct ggml_context * ctx);
int ggml_nelements(const struct ggml_tensor * tensor);
size_t ggml_nbytes (const struct ggml_tensor * tensor);
size_t ggml_type_size (enum ggml_type type);
int ggml_blck_size (enum ggml_type type);
size_t ggml_type_size (enum ggml_type type); // size in bytes for all elements in a block
float ggml_type_sizef(enum ggml_type type); // ggml_type_size()/ggml_blck_size() as float
size_t ggml_element_size(const struct ggml_tensor * tensor);
struct ggml_context * ggml_init(struct ggml_init_params params);
@@ -336,6 +343,9 @@ size_t ggml_used_mem(const struct ggml_context * ctx);
size_t ggml_set_scratch(struct ggml_context * ctx, struct ggml_scratch scratch);
bool ggml_mlock_supported(void);
bool ggml_mlock(struct ggml_context * ctx, char ** err_p);
struct ggml_tensor * ggml_new_tensor(
struct ggml_context * ctx,
enum ggml_type type,
@@ -466,12 +476,20 @@ struct ggml_tensor * ggml_gelu(
struct ggml_context * ctx,
struct ggml_tensor * a);
struct ggml_tensor * ggml_silu(
struct ggml_context * ctx,
struct ggml_tensor * a);
// normalize along rows
// TODO: eps is hardcoded to 1e-5 for now
struct ggml_tensor * ggml_norm(
struct ggml_context * ctx,
struct ggml_tensor * a);
struct ggml_tensor * ggml_rms_norm(
struct ggml_context * ctx,
struct ggml_tensor * a);
// A: m rows, n columns
// B: p rows, n columns (i.e. we transpose it internally)
// result is m columns, p rows
@@ -726,6 +744,13 @@ enum ggml_opt_result ggml_opt(
struct ggml_opt_params params,
struct ggml_tensor * f);
//
// quantization
//
size_t ggml_quantize_q4_0(const float * src, void * dst, int n, int k, int qk, int64_t * hist);
size_t ggml_quantize_q4_1(const float * src, void * dst, int n, int k, int qk, int64_t * hist);
//
// system info
//