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

Add changes to fix compiler issues

Browse source
This commit is contained in:
Srihari-mcw 2024-07-14 23:19:28 -07:00
parent db6657eeaf
commit f7ce132258
2 changed files with 7 additions and 34 deletions
Download patch file Download diff file Expand all files Collapse all files

6
ggml/src/ggml-quants.c
View file

@ -15279,7 +15279,6 @@ static bool validate_bf16(ggml_half f, size_t i) {
} \ } \
} }
<<<<<<< HEAD
#define VALIDATE_ROW_DATA_DVEC_F16_IMPL(type, data, nb, nr) \ #define VALIDATE_ROW_DATA_DVEC_F16_IMPL(type, data, nb, nr) \
const type * q = (const type *) (data); \ const type * q = (const type *) (data); \
for (size_t i = 0; i < (nb); ++i) { \ for (size_t i = 0; i < (nb); ++i) { \
@ -15287,13 +15286,14 @@ static bool validate_bf16(ggml_half f, size_t i) {
if (!validate_fp16(q[i].d[j], i)) { \ if (!validate_fp16(q[i].d[j], i)) { \
return false; \ return false; \
} \ } \
======= } \
}
#define VALIDATE_ROW_DATA_D_B16_IMPL(type, data, nb) \ #define VALIDATE_ROW_DATA_D_B16_IMPL(type, data, nb) \
const type * q = (const type *) (data); \ const type * q = (const type *) (data); \
for (size_t i = 0; i < (nb); ++i) { \ for (size_t i = 0; i < (nb); ++i) { \
if (!validate_bf16((q[i].d), i)) { \ if (!validate_bf16((q[i].d), i)) { \
return false; \ return false; \
>>>>>>> ed837022 (Introduce Q4_0 and Q8_0 quantizations with BF16 delta values)
} \ } \
} }

35
ggml/src/ggml.c
View file

@ -1041,7 +1041,7 @@ static const ggml_type_traits_t type_traits[GGML_TYPE_COUNT] = {
.is_quantized = true, .is_quantized = true,
.to_float = (ggml_to_float_t) dequantize_row_q4_0_b16, .to_float = (ggml_to_float_t) dequantize_row_q4_0_b16,
.from_float = quantize_row_q4_0_b16, .from_float = quantize_row_q4_0_b16,
.from_float_reference = (ggml_from_float_t) quantize_row_q4_0_b16_ref, .from_float_ref = (ggml_from_float_t) quantize_row_q4_0_b16_ref,
.vec_dot = ggml_vec_dot_q4_0_b16_q8_0_b16, .vec_dot = ggml_vec_dot_q4_0_b16_q8_0_b16,
.vec_dot_type = GGML_TYPE_Q8_0_B16, .vec_dot_type = GGML_TYPE_Q8_0_B16,
.nrows = 1, .nrows = 1,
@ -1053,7 +1053,7 @@ static const ggml_type_traits_t type_traits[GGML_TYPE_COUNT] = {
.is_quantized = true, .is_quantized = true,
.to_float = (ggml_to_float_t) dequantize_row_q8_0_b16, .to_float = (ggml_to_float_t) dequantize_row_q8_0_b16,
.from_float = quantize_row_q8_0_b16, .from_float = quantize_row_q8_0_b16,
.from_float_reference = (ggml_from_float_t) quantize_row_q8_0_b16_ref, .from_float_ref = (ggml_from_float_t) quantize_row_q8_0_b16_ref,
.vec_dot = ggml_vec_dot_q8_0_b16_q8_0_b16, .vec_dot = ggml_vec_dot_q8_0_b16_q8_0_b16,
.vec_dot_type = GGML_TYPE_Q8_0_B16, .vec_dot_type = GGML_TYPE_Q8_0_B16,
.nrows = 1, .nrows = 1,
@ -3334,15 +3334,16 @@ enum ggml_type ggml_ftype_to_ggml_type(enum ggml_ftype ftype) {
enum ggml_type wtype = GGML_TYPE_COUNT; enum ggml_type wtype = GGML_TYPE_COUNT;
switch (ftype) { switch (ftype) {
<<<<<<< HEAD
case GGML_FTYPE_ALL_F32: wtype = GGML_TYPE_F32; break; case GGML_FTYPE_ALL_F32: wtype = GGML_TYPE_F32; break;
case GGML_FTYPE_MOSTLY_F16: wtype = GGML_TYPE_F16; break; case GGML_FTYPE_MOSTLY_F16: wtype = GGML_TYPE_F16; break;
case GGML_FTYPE_MOSTLY_BF16: wtype = GGML_TYPE_BF16; break; case GGML_FTYPE_MOSTLY_BF16: wtype = GGML_TYPE_BF16; break;
case GGML_FTYPE_MOSTLY_Q4_0: wtype = GGML_TYPE_Q4_0; break; case GGML_FTYPE_MOSTLY_Q4_0: wtype = GGML_TYPE_Q4_0; break;
case GGML_FTYPE_MOSTLY_Q4_0_B16: wtype = GGML_TYPE_Q4_0_B16; break;
case GGML_FTYPE_MOSTLY_Q4_1: wtype = GGML_TYPE_Q4_1; break; case GGML_FTYPE_MOSTLY_Q4_1: wtype = GGML_TYPE_Q4_1; break;
case GGML_FTYPE_MOSTLY_Q5_0: wtype = GGML_TYPE_Q5_0; break; case GGML_FTYPE_MOSTLY_Q5_0: wtype = GGML_TYPE_Q5_0; break;
case GGML_FTYPE_MOSTLY_Q5_1: wtype = GGML_TYPE_Q5_1; break; case GGML_FTYPE_MOSTLY_Q5_1: wtype = GGML_TYPE_Q5_1; break;
case GGML_FTYPE_MOSTLY_Q8_0: wtype = GGML_TYPE_Q8_0; break; case GGML_FTYPE_MOSTLY_Q8_0: wtype = GGML_TYPE_Q8_0; break;
case GGML_FTYPE_MOSTLY_Q8_0_B16: wtype = GGML_TYPE_Q8_0_B16; break;
case GGML_FTYPE_MOSTLY_Q2_K: wtype = GGML_TYPE_Q2_K; break; case GGML_FTYPE_MOSTLY_Q2_K: wtype = GGML_TYPE_Q2_K; break;
case GGML_FTYPE_MOSTLY_Q3_K: wtype = GGML_TYPE_Q3_K; break; case GGML_FTYPE_MOSTLY_Q3_K: wtype = GGML_TYPE_Q3_K; break;
case GGML_FTYPE_MOSTLY_Q4_K: wtype = GGML_TYPE_Q4_K; break; case GGML_FTYPE_MOSTLY_Q4_K: wtype = GGML_TYPE_Q4_K; break;
@ -3362,34 +3363,6 @@ enum ggml_type ggml_ftype_to_ggml_type(enum ggml_ftype ftype) {
case GGML_FTYPE_MOSTLY_Q4_0_8_8: wtype = GGML_TYPE_Q4_0_8_8; break; case GGML_FTYPE_MOSTLY_Q4_0_8_8: wtype = GGML_TYPE_Q4_0_8_8; break;
case GGML_FTYPE_UNKNOWN: wtype = GGML_TYPE_COUNT; break; case GGML_FTYPE_UNKNOWN: wtype = GGML_TYPE_COUNT; break;
case GGML_FTYPE_MOSTLY_Q4_1_SOME_F16: wtype = GGML_TYPE_COUNT; break; case GGML_FTYPE_MOSTLY_Q4_1_SOME_F16: wtype = GGML_TYPE_COUNT; break;
=======
case GGML_FTYPE_ALL_F32: wtype = GGML_TYPE_F32; break;
case GGML_FTYPE_MOSTLY_F16: wtype = GGML_TYPE_F16; break;
case GGML_FTYPE_MOSTLY_BF16: wtype = GGML_TYPE_BF16; break;
case GGML_FTYPE_MOSTLY_Q4_0: wtype = GGML_TYPE_Q4_0; break;
case GGML_FTYPE_MOSTLY_Q4_0_B16: wtype = GGML_TYPE_Q4_0_B16; break;
case GGML_FTYPE_MOSTLY_Q4_1: wtype = GGML_TYPE_Q4_1; break;
case GGML_FTYPE_MOSTLY_Q5_0: wtype = GGML_TYPE_Q5_0; break;
case GGML_FTYPE_MOSTLY_Q5_1: wtype = GGML_TYPE_Q5_1; break;
case GGML_FTYPE_MOSTLY_Q8_0: wtype = GGML_TYPE_Q8_0; break;
case GGML_FTYPE_MOSTLY_Q8_0_B16: wtype = GGML_TYPE_Q8_0_B16; break;
case GGML_FTYPE_MOSTLY_Q2_K: wtype = GGML_TYPE_Q2_K; break;
case GGML_FTYPE_MOSTLY_Q3_K: wtype = GGML_TYPE_Q3_K; break;
case GGML_FTYPE_MOSTLY_Q4_K: wtype = GGML_TYPE_Q4_K; break;
case GGML_FTYPE_MOSTLY_Q5_K: wtype = GGML_TYPE_Q5_K; break;
case GGML_FTYPE_MOSTLY_Q6_K: wtype = GGML_TYPE_Q6_K; break;
case GGML_FTYPE_MOSTLY_IQ2_XXS: wtype = GGML_TYPE_IQ2_XXS; break;
case GGML_FTYPE_MOSTLY_IQ2_XS: wtype = GGML_TYPE_IQ2_XS; break;
case GGML_FTYPE_MOSTLY_IQ3_XXS: wtype = GGML_TYPE_IQ3_XXS; break;
case GGML_FTYPE_MOSTLY_IQ1_S: wtype = GGML_TYPE_IQ1_S; break;
case GGML_FTYPE_MOSTLY_IQ1_M: wtype = GGML_TYPE_IQ1_M; break;
case GGML_FTYPE_MOSTLY_IQ4_NL: wtype = GGML_TYPE_IQ4_NL; break;
case GGML_FTYPE_MOSTLY_IQ4_XS: wtype = GGML_TYPE_IQ4_XS; break;
case GGML_FTYPE_MOSTLY_IQ3_S: wtype = GGML_TYPE_IQ3_S; break;
case GGML_FTYPE_MOSTLY_IQ2_S: wtype = GGML_TYPE_IQ2_S; break;
case GGML_FTYPE_UNKNOWN: wtype = GGML_TYPE_COUNT; break;
case GGML_FTYPE_MOSTLY_Q4_1_SOME_F16: wtype = GGML_TYPE_COUNT; break;
>>>>>>> ed837022 (Introduce Q4_0 and Q8_0 quantizations with BF16 delta values)
} }
GGML_ASSERT(wtype != GGML_TYPE_COUNT); GGML_ASSERT(wtype != GGML_TYPE_COUNT);