vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Merge branch 'master' into jon/tall-and-skinny-matmul

Browse source
This commit is contained in:
jon-chuang 2023-04-16 02:10:58 +08:00 • committed by GitHub
commit b208c252a2
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
3 changed files with 505 additions and 358 deletions
Download patch file Download diff file Expand all files Collapse all files

2
Makefile
View file

@ -133,7 +133,7 @@ $(info I CC: $(CCV))
$(info I CXX: $(CXXV)) $(info I CXX: $(CXXV))
$(info ) $(info )
default: main quantize perplexity embedding default: main quantize quantize-stats perplexity embedding
# #
# Build library # Build library

859
ggml.c
View file

File diff suppressed because it is too large Load diff

2
ggml.h
View file

@ -204,6 +204,7 @@ enum ggml_type {
GGML_TYPE_F16 = 1, GGML_TYPE_F16 = 1,
GGML_TYPE_Q4_0 = 2, GGML_TYPE_Q4_0 = 2,
GGML_TYPE_Q4_1 = 3, GGML_TYPE_Q4_1 = 3,
GGML_TYPE_Q8_0 = 4,
GGML_TYPE_I8, GGML_TYPE_I8,
GGML_TYPE_I16, GGML_TYPE_I16,
GGML_TYPE_I32, GGML_TYPE_I32,
@ -836,6 +837,7 @@ typedef struct {
dequantize_row_q_t dequantize_row_q; dequantize_row_q_t dequantize_row_q;
quantize_row_q_t quantize_row_q; quantize_row_q_t quantize_row_q;
quantize_row_q_t quantize_row_q_reference; quantize_row_q_t quantize_row_q_reference;
quantize_row_q_t quantize_row_q_dot;
vec_dot_q_t vec_dot_q; vec_dot_q_t vec_dot_q;
} quantize_fns_t; } quantize_fns_t;