diff options
Diffstat (limited to 'ggml/src/ggml.c')
| -rw-r--r-- | ggml/src/ggml.c | 30 |
1 files changed, 26 insertions, 4 deletions
diff --git a/ggml/src/ggml.c b/ggml/src/ggml.c index 97fa81b1..a9f795ae 100644 --- a/ggml/src/ggml.c +++ b/ggml/src/ggml.c @@ -920,8 +920,8 @@ static const ggml_type_traits_t type_traits[GGML_TYPE_COUNT] = { .type_size = sizeof(block_iq2_xxs), .is_quantized = true, .to_float = (ggml_to_float_t) dequantize_row_iq2_xxs, - .from_float = NULL, - .from_float_ref = NULL, + .from_float = quantize_row_iq2_xxs, + .from_float_ref = (ggml_from_float_t)quantize_row_iq2_xxs_ref, .vec_dot = ggml_vec_dot_iq2_xxs_q8_K, .vec_dot_type = GGML_TYPE_Q8_K, .nrows = 1, @@ -933,8 +933,8 @@ static const ggml_type_traits_t type_traits[GGML_TYPE_COUNT] = { .type_size = sizeof(block_iq2_xs), .is_quantized = true, .to_float = (ggml_to_float_t) dequantize_row_iq2_xs, - .from_float = NULL, - .from_float_ref = NULL, + .from_float = quantize_row_iq2_xs, + .from_float_ref = (ggml_from_float_t)quantize_row_iq2_xs_ref, .vec_dot = ggml_vec_dot_iq2_xs_q8_K, .vec_dot_type = GGML_TYPE_Q8_K, .nrows = 1, @@ -1193,6 +1193,19 @@ static const ggml_type_traits_t type_traits[GGML_TYPE_COUNT] = { .nrows = 1, .row_meta_size = 0, }, + [GGML_TYPE_IQ2_KS] = { + .type_name = "iq2_ks", + .blck_size = QK_K, + .type_size = sizeof(block_iq2_ks), + .is_quantized = true, + .to_float = (ggml_to_float_t) dequantize_row_iq2_ks, + .from_float = quantize_row_iq2_ks, + .from_float_ref = (ggml_from_float_t)quantize_row_iq2_ks_ref, + .vec_dot = vec_dot_iq2_ks_q8_k, + .vec_dot_type = GGML_TYPE_Q8_K, + .nrows = 1, + .row_meta_size = 2, + }, [GGML_TYPE_IQ3_K] = { .type_name = "iq3_k", .blck_size = QK_K, @@ -3906,6 +3919,7 @@ enum ggml_type ggml_ftype_to_ggml_type(enum ggml_ftype ftype) { case GGML_FTYPE_MOSTLY_IQ4_XS: wtype = GGML_TYPE_IQ4_XS; break; case GGML_FTYPE_MOSTLY_IQ4_KS: wtype = GGML_TYPE_IQ4_KS; break; case GGML_FTYPE_MOSTLY_IQ2_K: wtype = GGML_TYPE_IQ2_K; break; + case GGML_FTYPE_MOSTLY_IQ2_KS: wtype = GGML_TYPE_IQ2_KS; break; case GGML_FTYPE_MOSTLY_IQ3_K: wtype = GGML_TYPE_IQ3_K; break; case GGML_FTYPE_MOSTLY_IQ4_K: wtype = GGML_TYPE_IQ4_K; break; case GGML_FTYPE_MOSTLY_IQ5_K: wtype = GGML_TYPE_IQ5_K; break; @@ -10406,6 +10420,7 @@ static void ggml_compute_forward_add( case GGML_TYPE_IQ4_XS: case GGML_TYPE_IQ4_KS: case GGML_TYPE_IQ2_K: + case GGML_TYPE_IQ2_KS: case GGML_TYPE_IQ3_K: case GGML_TYPE_IQ4_K: case GGML_TYPE_IQ5_K: @@ -10795,6 +10810,7 @@ static void ggml_compute_forward_add1( case GGML_TYPE_IQ4_XS: case GGML_TYPE_IQ4_KS: case GGML_TYPE_IQ2_K: + case GGML_TYPE_IQ2_KS: case GGML_TYPE_IQ3_K: case GGML_TYPE_IQ4_K: case GGML_TYPE_IQ5_K: @@ -10934,6 +10950,7 @@ static void ggml_compute_forward_acc( case GGML_TYPE_IQ4_XS: case GGML_TYPE_IQ4_KS: case GGML_TYPE_IQ2_K: + case GGML_TYPE_IQ2_KS: case GGML_TYPE_IQ3_K: case GGML_TYPE_IQ4_K: case GGML_TYPE_IQ5_K: @@ -14119,6 +14136,7 @@ static void ggml_compute_forward_out_prod( case GGML_TYPE_IQ4_XS: case GGML_TYPE_IQ4_KS: case GGML_TYPE_IQ2_K: + case GGML_TYPE_IQ2_KS: case GGML_TYPE_IQ3_K: case GGML_TYPE_IQ4_K: case GGML_TYPE_IQ5_K: @@ -14498,6 +14516,7 @@ static void ggml_compute_forward_set( case GGML_TYPE_IQ4_XS: case GGML_TYPE_IQ4_KS: case GGML_TYPE_IQ2_K: + case GGML_TYPE_IQ2_KS: case GGML_TYPE_IQ3_K: case GGML_TYPE_IQ4_K: case GGML_TYPE_IQ5_K: @@ -14771,6 +14790,7 @@ static void ggml_compute_forward_get_rows( case GGML_TYPE_IQ4_XS: case GGML_TYPE_IQ4_KS: case GGML_TYPE_IQ2_K: + case GGML_TYPE_IQ2_KS: case GGML_TYPE_IQ3_K: case GGML_TYPE_IQ4_K: case GGML_TYPE_IQ5_K: @@ -15371,6 +15391,7 @@ static void ggml_compute_forward_clamp( case GGML_TYPE_IQ4_XS: case GGML_TYPE_IQ4_KS: case GGML_TYPE_IQ2_K: + case GGML_TYPE_IQ2_KS: case GGML_TYPE_IQ3_K: case GGML_TYPE_IQ4_K: case GGML_TYPE_IQ5_K: @@ -22188,6 +22209,7 @@ size_t ggml_quantize_chunk( case GGML_TYPE_IQ4_XS: result = quantize_iq4_xs (src + start, (char *) dst + start_row * row_size, nrows, n_per_row, imatrix); break; case GGML_TYPE_IQ4_KS: result = quantize_iq4_ks (src + start, (char *) dst + start_row * row_size, nrows, n_per_row, imatrix); break; case GGML_TYPE_IQ2_K: result = quantize_iq2_k (src + start, (char *) dst + start_row * row_size, nrows, n_per_row, imatrix); break; + case GGML_TYPE_IQ2_KS: result = quantize_iq2_ks (src + start, (char *) dst + start_row * row_size, nrows, n_per_row, imatrix); break; case GGML_TYPE_IQ3_K: result = quantize_iq3_k (src + start, (char *) dst + start_row * row_size, nrows, n_per_row, imatrix); break; case GGML_TYPE_IQ4_K: result = quantize_iq4_k (src + start, (char *) dst + start_row * row_size, nrows, n_per_row, imatrix); break; case GGML_TYPE_IQ5_K: result = quantize_iq5_k (src + start, (char *) dst + start_row * row_size, nrows, n_per_row, imatrix); break; |