ggml : minor naming changes (#8433)
* ggml : minor naming changes ggml-ci * ggml : use PRId64 [no ci] * ggml : revert FA K/Q names
This commit is contained in:
Georgi Gerganov
GitHub
parent
b549a1bbef
commit
370b1f7e7a
10 changed files with 192 additions and 189 deletions
|
|
@ -658,7 +658,7 @@ static inline __m128i packNibbles( __m256i bytes ) {
|
|||
#endif //__loongarch_asx
|
||||
|
||||
// reference implementation for deterministic creation of model files
|
||||
void quantize_row_q4_0_reference(const float * restrict x, block_q4_0 * restrict y, int64_t k) {
|
||||
void quantize_row_q4_0_ref(const float * restrict x, block_q4_0 * restrict y, int64_t k) {
|
||||
static const int qk = QK4_0;
|
||||
|
||||
assert(k % qk == 0);
|
||||
|
|
@ -696,11 +696,11 @@ void quantize_row_q4_0_reference(const float * restrict x, block_q4_0 * restrict
|
|||
}
|
||||
|
||||
void quantize_row_q4_0(const float * restrict x, void * restrict y, int64_t k) {
|
||||
quantize_row_q4_0_reference(x, y, k);
|
||||
quantize_row_q4_0_ref(x, y, k);
|
||||
}
|
||||
|
||||
|
||||
void quantize_row_q4_1_reference(const float * restrict x, block_q4_1 * restrict y, int64_t k) {
|
||||
void quantize_row_q4_1_ref(const float * restrict x, block_q4_1 * restrict y, int64_t k) {
|
||||
const int qk = QK4_1;
|
||||
|
||||
assert(k % qk == 0);
|
||||
|
|
@ -738,10 +738,10 @@ void quantize_row_q4_1_reference(const float * restrict x, block_q4_1 * restrict
|
|||
}
|
||||
|
||||
void quantize_row_q4_1(const float * restrict x, void * restrict y, int64_t k) {
|
||||
quantize_row_q4_1_reference(x, y, k);
|
||||
quantize_row_q4_1_ref(x, y, k);
|
||||
}
|
||||
|
||||
void quantize_row_q5_0_reference(const float * restrict x, block_q5_0 * restrict y, int64_t k) {
|
||||
void quantize_row_q5_0_ref(const float * restrict x, block_q5_0 * restrict y, int64_t k) {
|
||||
static const int qk = QK5_0;
|
||||
|
||||
assert(k % qk == 0);
|
||||
|
|
@ -786,10 +786,10 @@ void quantize_row_q5_0_reference(const float * restrict x, block_q5_0 * restrict
|
|||
}
|
||||
|
||||
void quantize_row_q5_0(const float * restrict x, void * restrict y, int64_t k) {
|
||||
quantize_row_q5_0_reference(x, y, k);
|
||||
quantize_row_q5_0_ref(x, y, k);
|
||||
}
|
||||
|
||||
void quantize_row_q5_1_reference(const float * restrict x, block_q5_1 * restrict y, int64_t k) {
|
||||
void quantize_row_q5_1_ref(const float * restrict x, block_q5_1 * restrict y, int64_t k) {
|
||||
const int qk = QK5_1;
|
||||
|
||||
assert(k % qk == 0);
|
||||
|
|
@ -834,11 +834,11 @@ void quantize_row_q5_1_reference(const float * restrict x, block_q5_1 * restrict
|
|||
}
|
||||
|
||||
void quantize_row_q5_1(const float * restrict x, void * restrict y, int64_t k) {
|
||||
quantize_row_q5_1_reference(x, y, k);
|
||||
quantize_row_q5_1_ref(x, y, k);
|
||||
}
|
||||
|
||||
// reference implementation for deterministic creation of model files
|
||||
void quantize_row_q8_0_reference(const float * restrict x, block_q8_0 * restrict y, int64_t k) {
|
||||
void quantize_row_q8_0_ref(const float * restrict x, block_q8_0 * restrict y, int64_t k) {
|
||||
assert(k % QK8_0 == 0);
|
||||
const int nb = k / QK8_0;
|
||||
|
||||
|
|
@ -1144,12 +1144,12 @@ void quantize_row_q8_0(const float * restrict x, void * restrict vy, int64_t k)
|
|||
#else
|
||||
GGML_UNUSED(nb);
|
||||
// scalar
|
||||
quantize_row_q8_0_reference(x, y, k);
|
||||
quantize_row_q8_0_ref(x, y, k);
|
||||
#endif
|
||||
}
|
||||
|
||||
// reference implementation for deterministic creation of model files
|
||||
void quantize_row_q8_1_reference(const float * restrict x, block_q8_1 * restrict y, int64_t k) {
|
||||
void quantize_row_q8_1_ref(const float * restrict x, block_q8_1 * restrict y, int64_t k) {
|
||||
assert(QK8_1 == 32);
|
||||
assert(k % QK8_1 == 0);
|
||||
const int nb = k / QK8_1;
|
||||
|
|
@ -1508,7 +1508,7 @@ void quantize_row_q8_1(const float * restrict x, void * restrict vy, int64_t k)
|
|||
#else
|
||||
GGML_UNUSED(nb);
|
||||
// scalar
|
||||
quantize_row_q8_1_reference(x, y, k);
|
||||
quantize_row_q8_1_ref(x, y, k);
|
||||
#endif
|
||||
}
|
||||
|
||||
|
|
@ -1899,7 +1899,7 @@ static inline void get_scale_min_k4(int j, const uint8_t * restrict q, uint8_t *
|
|||
|
||||
//========================- 2-bit (de)-quantization
|
||||
|
||||
void quantize_row_q2_K_reference(const float * restrict x, block_q2_K * restrict y, int64_t k) {
|
||||
void quantize_row_q2_K_ref(const float * restrict x, block_q2_K * restrict y, int64_t k) {
|
||||
assert(k % QK_K == 0);
|
||||
const int nb = k / QK_K;
|
||||
|
||||
|
|
@ -2002,7 +2002,7 @@ void dequantize_row_q2_K(const block_q2_K * restrict x, float * restrict y, int6
|
|||
}
|
||||
|
||||
void quantize_row_q2_K(const float * restrict x, void * restrict vy, int64_t k) {
|
||||
quantize_row_q2_K_reference(x, vy, k);
|
||||
quantize_row_q2_K_ref(x, vy, k);
|
||||
}
|
||||
|
||||
static float make_qkx3_quants(int n, int nmax, const float * restrict x, const float * restrict weights,
|
||||
|
|
@ -2226,7 +2226,7 @@ static void quantize_row_q2_K_impl(const float * restrict x, block_q2_K * restri
|
|||
size_t quantize_q2_K(const float * restrict src, void * restrict dst, int64_t nrow, int64_t n_per_row, const float * quant_weights) {
|
||||
size_t row_size = ggml_row_size(GGML_TYPE_Q2_K, n_per_row);
|
||||
if (!quant_weights) {
|
||||
quantize_row_q2_K_reference(src, dst, (int64_t)nrow*n_per_row);
|
||||
quantize_row_q2_K_ref(src, dst, (int64_t)nrow*n_per_row);
|
||||
}
|
||||
else {
|
||||
char * qrow = (char *)dst;
|
||||
|
|
@ -2241,7 +2241,7 @@ size_t quantize_q2_K(const float * restrict src, void * restrict dst, int64_t nr
|
|||
|
||||
//========================= 3-bit (de)-quantization
|
||||
|
||||
void quantize_row_q3_K_reference(const float * restrict x, block_q3_K * restrict y, int64_t k) {
|
||||
void quantize_row_q3_K_ref(const float * restrict x, block_q3_K * restrict y, int64_t k) {
|
||||
assert(k % QK_K == 0);
|
||||
const int nb = k / QK_K;
|
||||
|
||||
|
|
@ -2368,7 +2368,7 @@ void dequantize_row_q3_K(const block_q3_K * restrict x, float * restrict y, int6
|
|||
}
|
||||
|
||||
void quantize_row_q3_K(const float * restrict x, void * restrict vy, int64_t k) {
|
||||
quantize_row_q3_K_reference(x, vy, k);
|
||||
quantize_row_q3_K_ref(x, vy, k);
|
||||
}
|
||||
|
||||
static void quantize_row_q3_K_impl(const float * restrict x, block_q3_K * restrict y, int64_t n_per_row, const float * restrict quant_weights) {
|
||||
|
|
@ -2458,7 +2458,7 @@ static void quantize_row_q3_K_impl(const float * restrict x, block_q3_K * restri
|
|||
size_t quantize_q3_K(const float * restrict src, void * restrict dst, int64_t nrow, int64_t n_per_row, const float * quant_weights) {
|
||||
size_t row_size = ggml_row_size(GGML_TYPE_Q3_K, n_per_row);
|
||||
if (!quant_weights) {
|
||||
quantize_row_q3_K_reference(src, dst, (int64_t)nrow*n_per_row);
|
||||
quantize_row_q3_K_ref(src, dst, (int64_t)nrow*n_per_row);
|
||||
}
|
||||
else {
|
||||
char * qrow = (char *)dst;
|
||||
|
|
@ -2473,7 +2473,7 @@ size_t quantize_q3_K(const float * restrict src, void * restrict dst, int64_t nr
|
|||
|
||||
// ====================== 4-bit (de)-quantization
|
||||
|
||||
void quantize_row_q4_K_reference(const float * restrict x, block_q4_K * restrict y, int64_t k) {
|
||||
void quantize_row_q4_K_ref(const float * restrict x, block_q4_K * restrict y, int64_t k) {
|
||||
assert(k % QK_K == 0);
|
||||
const int nb = k / QK_K;
|
||||
|
||||
|
|
@ -2572,7 +2572,7 @@ void dequantize_row_q4_K(const block_q4_K * restrict x, float * restrict y, int6
|
|||
void quantize_row_q4_K(const float * restrict x, void * restrict vy, int64_t k) {
|
||||
assert(k % QK_K == 0);
|
||||
block_q4_K * restrict y = vy;
|
||||
quantize_row_q4_K_reference(x, y, k);
|
||||
quantize_row_q4_K_ref(x, y, k);
|
||||
}
|
||||
|
||||
static void quantize_row_q4_K_impl(const float * restrict x, block_q4_K * restrict y, int64_t n_per_row, const float * quant_weights) {
|
||||
|
|
@ -2651,7 +2651,7 @@ static void quantize_row_q4_K_impl(const float * restrict x, block_q4_K * restri
|
|||
size_t quantize_q4_K(const float * restrict src, void * restrict dst, int64_t nrow, int64_t n_per_row, const float * quant_weights) {
|
||||
size_t row_size = ggml_row_size(GGML_TYPE_Q4_K, n_per_row);
|
||||
if (!quant_weights) {
|
||||
quantize_row_q4_K_reference(src, dst, (int64_t)nrow*n_per_row);
|
||||
quantize_row_q4_K_ref(src, dst, (int64_t)nrow*n_per_row);
|
||||
}
|
||||
else {
|
||||
char * qrow = (char *)dst;
|
||||
|
|
@ -2666,7 +2666,7 @@ size_t quantize_q4_K(const float * restrict src, void * restrict dst, int64_t nr
|
|||
|
||||
// ====================== 5-bit (de)-quantization
|
||||
|
||||
void quantize_row_q5_K_reference(const float * restrict x, block_q5_K * restrict y, int64_t k) {
|
||||
void quantize_row_q5_K_ref(const float * restrict x, block_q5_K * restrict y, int64_t k) {
|
||||
assert(k % QK_K == 0);
|
||||
const int64_t nb = k / QK_K;
|
||||
|
||||
|
|
@ -2783,7 +2783,7 @@ void dequantize_row_q5_K(const block_q5_K * restrict x, float * restrict y, int6
|
|||
void quantize_row_q5_K(const float * restrict x, void * restrict vy, int64_t k) {
|
||||
assert(k % QK_K == 0);
|
||||
block_q5_K * restrict y = vy;
|
||||
quantize_row_q5_K_reference(x, y, k);
|
||||
quantize_row_q5_K_ref(x, y, k);
|
||||
}
|
||||
|
||||
static void quantize_row_q5_K_impl(const float * restrict x, block_q5_K * restrict y, int64_t n_per_row, const float * quant_weights) {
|
||||
|
|
@ -2882,7 +2882,7 @@ static void quantize_row_q5_K_impl(const float * restrict x, block_q5_K * restri
|
|||
size_t quantize_q5_K(const float * restrict src, void * restrict dst, int64_t nrow, int64_t n_per_row, const float * quant_weights) {
|
||||
size_t row_size = ggml_row_size(GGML_TYPE_Q5_K, n_per_row);
|
||||
if (!quant_weights) {
|
||||
quantize_row_q5_K_reference(src, dst, (int64_t)nrow*n_per_row);
|
||||
quantize_row_q5_K_ref(src, dst, (int64_t)nrow*n_per_row);
|
||||
}
|
||||
else {
|
||||
char * qrow = (char *)dst;
|
||||
|
|
@ -2897,7 +2897,7 @@ size_t quantize_q5_K(const float * restrict src, void * restrict dst, int64_t nr
|
|||
|
||||
// ====================== 6-bit (de)-quantization
|
||||
|
||||
void quantize_row_q6_K_reference(const float * restrict x, block_q6_K * restrict y, int64_t k) {
|
||||
void quantize_row_q6_K_ref(const float * restrict x, block_q6_K * restrict y, int64_t k) {
|
||||
assert(k % QK_K == 0);
|
||||
const int64_t nb = k / QK_K;
|
||||
|
||||
|
|
@ -3001,7 +3001,7 @@ void dequantize_row_q6_K(const block_q6_K * restrict x, float * restrict y, int6
|
|||
void quantize_row_q6_K(const float * restrict x, void * restrict vy, int64_t k) {
|
||||
assert(k % QK_K == 0);
|
||||
block_q6_K * restrict y = vy;
|
||||
quantize_row_q6_K_reference(x, y, k);
|
||||
quantize_row_q6_K_ref(x, y, k);
|
||||
}
|
||||
|
||||
static void quantize_row_q6_K_impl(const float * restrict x, block_q6_K * restrict y, int64_t n_per_row, const float * quant_weights) {
|
||||
|
|
@ -3091,7 +3091,7 @@ static void quantize_row_q6_K_impl(const float * restrict x, block_q6_K * restri
|
|||
size_t quantize_q6_K(const float * restrict src, void * restrict dst, int64_t nrow, int64_t n_per_row, const float * quant_weights) {
|
||||
size_t row_size = ggml_row_size(GGML_TYPE_Q6_K, n_per_row);
|
||||
if (!quant_weights) {
|
||||
quantize_row_q6_K_reference(src, dst, (int64_t)nrow*n_per_row);
|
||||
quantize_row_q6_K_ref(src, dst, (int64_t)nrow*n_per_row);
|
||||
}
|
||||
else {
|
||||
char * qrow = (char *)dst;
|
||||
|
|
@ -3108,7 +3108,7 @@ static void quantize_row_q4_0_impl(const float * restrict x, block_q4_0 * restri
|
|||
static_assert(QK4_0 == 32, "QK4_0 must be 32");
|
||||
|
||||
if (!quant_weights) {
|
||||
quantize_row_q4_0_reference(x, y, n_per_row);
|
||||
quantize_row_q4_0_ref(x, y, n_per_row);
|
||||
return;
|
||||
}
|
||||
|
||||
|
|
@ -3134,7 +3134,7 @@ static void quantize_row_q4_0_impl(const float * restrict x, block_q4_0 * restri
|
|||
|
||||
size_t quantize_q4_0(const float * restrict src, void * restrict dst, int64_t nrow, int64_t n_per_row, const float * quant_weights) {
|
||||
if (!quant_weights) {
|
||||
quantize_row_q4_0_reference(src, dst, (int64_t)nrow*n_per_row);
|
||||
quantize_row_q4_0_ref(src, dst, (int64_t)nrow*n_per_row);
|
||||
return nrow * ggml_row_size(GGML_TYPE_Q4_0, n_per_row);
|
||||
}
|
||||
size_t row_size = ggml_row_size(GGML_TYPE_Q4_0, n_per_row);
|
||||
|
|
@ -3151,7 +3151,7 @@ static void quantize_row_q4_1_impl(const float * restrict x, block_q4_1 * restri
|
|||
static_assert(QK4_1 == 32, "QK4_1 must be 32");
|
||||
|
||||
if (!quant_weights) {
|
||||
quantize_row_q4_1_reference(x, y, n_per_row);
|
||||
quantize_row_q4_1_ref(x, y, n_per_row);
|
||||
return;
|
||||
}
|
||||
|
||||
|
|
@ -3179,7 +3179,7 @@ static void quantize_row_q4_1_impl(const float * restrict x, block_q4_1 * restri
|
|||
|
||||
size_t quantize_q4_1(const float * restrict src, void * restrict dst, int64_t nrow, int64_t n_per_row, const float * quant_weights) {
|
||||
if (!quant_weights) {
|
||||
quantize_row_q4_1_reference(src, dst, (int64_t)nrow*n_per_row);
|
||||
quantize_row_q4_1_ref(src, dst, (int64_t)nrow*n_per_row);
|
||||
return nrow * ggml_row_size(GGML_TYPE_Q4_1, n_per_row);
|
||||
}
|
||||
size_t row_size = ggml_row_size(GGML_TYPE_Q4_1, n_per_row);
|
||||
|
|
@ -3196,7 +3196,7 @@ static void quantize_row_q5_0_impl(const float * restrict x, block_q5_0 * restri
|
|||
static_assert(QK5_0 == 32, "QK5_0 must be 32");
|
||||
|
||||
if (!quant_weights) {
|
||||
quantize_row_q5_0_reference(x, y, n_per_row);
|
||||
quantize_row_q5_0_ref(x, y, n_per_row);
|
||||
return;
|
||||
}
|
||||
|
||||
|
|
@ -3233,7 +3233,7 @@ static void quantize_row_q5_0_impl(const float * restrict x, block_q5_0 * restri
|
|||
|
||||
size_t quantize_q5_0(const float * restrict src, void * restrict dst, int64_t nrow, int64_t n_per_row, const float * quant_weights) {
|
||||
if (!quant_weights) {
|
||||
quantize_row_q5_0_reference(src, dst, (int64_t)nrow*n_per_row);
|
||||
quantize_row_q5_0_ref(src, dst, (int64_t)nrow*n_per_row);
|
||||
return nrow * ggml_row_size(GGML_TYPE_Q5_0, n_per_row);
|
||||
}
|
||||
size_t row_size = ggml_row_size(GGML_TYPE_Q5_0, n_per_row);
|
||||
|
|
@ -3250,7 +3250,7 @@ static void quantize_row_q5_1_impl(const float * restrict x, block_q5_1 * restri
|
|||
static_assert(QK5_1 == 32, "QK5_1 must be 32");
|
||||
|
||||
if (!quant_weights) {
|
||||
quantize_row_q5_1_reference(x, y, n_per_row);
|
||||
quantize_row_q5_1_ref(x, y, n_per_row);
|
||||
return;
|
||||
}
|
||||
|
||||
|
|
@ -3286,7 +3286,7 @@ static void quantize_row_q5_1_impl(const float * restrict x, block_q5_1 * restri
|
|||
|
||||
size_t quantize_q5_1(const float * restrict src, void * restrict dst, int64_t nrow, int64_t n_per_row, const float * quant_weights) {
|
||||
if (!quant_weights) {
|
||||
quantize_row_q5_1_reference(src, dst, (int64_t)nrow*n_per_row);
|
||||
quantize_row_q5_1_ref(src, dst, (int64_t)nrow*n_per_row);
|
||||
return nrow * ggml_row_size(GGML_TYPE_Q5_1, n_per_row);
|
||||
}
|
||||
size_t row_size = ggml_row_size(GGML_TYPE_Q5_1, n_per_row);
|
||||
|
|
@ -3302,7 +3302,7 @@ size_t quantize_q5_1(const float * restrict src, void * restrict dst, int64_t nr
|
|||
size_t quantize_q8_0(const float * restrict src, void * restrict dst, int64_t nrow, int64_t n_per_row, const float * quant_weights) {
|
||||
(void)quant_weights; // not used
|
||||
const size_t row_size = ggml_row_size(GGML_TYPE_Q8_0, n_per_row);
|
||||
quantize_row_q8_0_reference(src, dst, (int64_t)nrow*n_per_row);
|
||||
quantize_row_q8_0_ref(src, dst, (int64_t)nrow*n_per_row);
|
||||
return nrow * row_size;
|
||||
}
|
||||
|
||||
|
|
@ -3590,7 +3590,7 @@ void dequantize_row_iq4_xs(const block_iq4_xs * restrict x, float * restrict y,
|
|||
|
||||
//===================================== Q8_K ==============================================
|
||||
|
||||
void quantize_row_q8_K_reference(const float * restrict x, block_q8_K * restrict y, int64_t k) {
|
||||
void quantize_row_q8_K_ref(const float * restrict x, block_q8_K * restrict y, int64_t k) {
|
||||
assert(k % QK_K == 0);
|
||||
const int64_t nb = k / QK_K;
|
||||
|
||||
|
|
@ -3641,7 +3641,7 @@ void dequantize_row_q8_K(const block_q8_K * restrict x, float * restrict y, int6
|
|||
}
|
||||
|
||||
void quantize_row_q8_K(const float * restrict x, void * restrict y, int64_t k) {
|
||||
quantize_row_q8_K_reference(x, y, k);
|
||||
quantize_row_q8_K_ref(x, y, k);
|
||||
}
|
||||
|
||||
//===================================== Dot ptoducts =================================
|
||||
|
|
@ -13530,10 +13530,10 @@ size_t quantize_iq3_xxs(const float * restrict src, void * restrict dst, int64_t
|
|||
void quantize_row_iq3_xxs(const float * restrict x, void * restrict vy, int64_t k) {
|
||||
assert(k % QK_K == 0);
|
||||
block_iq3_xxs * restrict y = vy;
|
||||
quantize_row_iq3_xxs_reference(x, y, k);
|
||||
quantize_row_iq3_xxs_ref(x, y, k);
|
||||
}
|
||||
|
||||
void quantize_row_iq3_xxs_reference(const float * restrict x, block_iq3_xxs * restrict y, int64_t k) {
|
||||
void quantize_row_iq3_xxs_ref(const float * restrict x, block_iq3_xxs * restrict y, int64_t k) {
|
||||
assert(k % QK_K == 0);
|
||||
quantize_row_iq3_xxs_impl(256, x, y, k, NULL);
|
||||
}
|
||||
|
|
@ -13746,10 +13746,10 @@ size_t quantize_iq3_s(const float * restrict src, void * restrict dst, int64_t n
|
|||
void quantize_row_iq3_s(const float * restrict x, void * restrict vy, int64_t k) {
|
||||
assert(k % QK_K == 0);
|
||||
block_iq3_s * restrict y = vy;
|
||||
quantize_row_iq3_s_reference(x, y, k);
|
||||
quantize_row_iq3_s_ref(x, y, k);
|
||||
}
|
||||
|
||||
void quantize_row_iq3_s_reference(const float * restrict x, block_iq3_s * restrict y, int64_t k) {
|
||||
void quantize_row_iq3_s_ref(const float * restrict x, block_iq3_s * restrict y, int64_t k) {
|
||||
assert(k % QK_K == 0);
|
||||
quantize_iq3_s(x, y, 1, k, NULL);
|
||||
}
|
||||
|
|
@ -14487,7 +14487,7 @@ void quantize_row_iq4_nl(const float * restrict x, void * restrict vy, int64_t k
|
|||
}
|
||||
}
|
||||
|
||||
void quantize_row_iq4_nl_reference(const float * restrict x, block_iq4_nl * restrict y, int64_t k) {
|
||||
void quantize_row_iq4_nl_ref(const float * restrict x, block_iq4_nl * restrict y, int64_t k) {
|
||||
assert(k % QK4_NL == 0);
|
||||
quantize_row_iq4_nl(x, y, k);
|
||||
}
|
||||
|
|
@ -14515,10 +14515,10 @@ size_t quantize_iq4_xs(const float * restrict src, void * restrict dst, int64_t
|
|||
void quantize_row_iq4_xs(const float * restrict x, void * restrict vy, int64_t k) {
|
||||
assert(k % QK_K == 0);
|
||||
block_iq4_xs * restrict y = vy;
|
||||
quantize_row_iq4_xs_reference(x, y, k);
|
||||
quantize_row_iq4_xs_ref(x, y, k);
|
||||
}
|
||||
|
||||
void quantize_row_iq4_xs_reference(const float * restrict x, block_iq4_xs * restrict y, int64_t k) {
|
||||
void quantize_row_iq4_xs_ref(const float * restrict x, block_iq4_xs * restrict y, int64_t k) {
|
||||
assert(k % QK_K == 0);
|
||||
quantize_iq4_xs(x, y, 1, k, NULL);
|
||||
}
|
||||
|
|
@ -14705,7 +14705,7 @@ size_t quantize_iq2_s(const float * restrict src, void * restrict dst, int64_t n
|
|||
return nrow * nblock * sizeof(block_iq2_s);
|
||||
}
|
||||
|
||||
void quantize_row_iq2_s_reference(const float * restrict x, block_iq2_s * restrict y, int64_t k) {
|
||||
void quantize_row_iq2_s_ref(const float * restrict x, block_iq2_s * restrict y, int64_t k) {
|
||||
assert(k % QK_K == 0);
|
||||
quantize_iq2_s(x, y, 1, k, NULL);
|
||||
}
|
||||
|
|
@ -14713,7 +14713,7 @@ void quantize_row_iq2_s_reference(const float * restrict x, block_iq2_s * restri
|
|||
void quantize_row_iq2_s(const float * restrict x, void * restrict vy, int64_t k) {
|
||||
assert(k % QK_K == 0);
|
||||
block_iq2_s * restrict y = vy;
|
||||
quantize_row_iq2_s_reference(x, y, k);
|
||||
quantize_row_iq2_s_ref(x, y, k);
|
||||
}
|
||||
|
||||
static bool validate_float(float f, size_t i) {
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue