Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-19 09:15:18 +02:00
Code Issues Packages Projects Releases Wiki Activity

interim merge. do not use

Browse Source
This commit is contained in:
Concedo
2023-05-13 00:33:55 +08:00
parent e9caff1cda
commit 08810d5fee
3 changed files with 188 additions and 10756 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml.c
+18
View File
@@ -393,6 +393,8 @@ void ggml_fp32_to_fp16_row(const float * x, ggml_fp16_t * y, size_t n) {
}
}
//legacy functions
#include "ggml_v2.c"
//
// timing
@@ -3064,6 +3066,8 @@ static const int GGML_BLCK_SIZE[GGML_TYPE_COUNT] = {
[GGML_TYPE_F16] = 1,
[GGML_TYPE_Q4_0] = QK4_0,
[GGML_TYPE_Q4_1] = QK4_1,
[GGML_TYPE_Q4_2] = QK4_2,
[GGML_TYPE_Q4_3] = QK4_3,
[GGML_TYPE_Q5_0] = QK5_0,
[GGML_TYPE_Q5_1] = QK5_1,
[GGML_TYPE_Q8_0] = QK8_0,
@@ -3079,6 +3083,8 @@ static const size_t GGML_TYPE_SIZE[GGML_TYPE_COUNT] = {
[GGML_TYPE_F16] = sizeof(ggml_fp16_t),
[GGML_TYPE_Q4_0] = sizeof(block_q4_0),
[GGML_TYPE_Q4_1] = sizeof(block_q4_1),
[GGML_TYPE_Q4_2] = sizeof(block_q4_2),
[GGML_TYPE_Q4_3] = sizeof(block_q4_3),
[GGML_TYPE_Q5_0] = sizeof(block_q5_0),
[GGML_TYPE_Q5_1] = sizeof(block_q5_1),
[GGML_TYPE_Q8_0] = sizeof(block_q8_0),
@@ -3095,6 +3101,8 @@ static const char * GGML_TYPE_NAME[GGML_TYPE_COUNT] = {
[GGML_TYPE_F16] = "f16",
[GGML_TYPE_Q4_0] = "q4_0",
[GGML_TYPE_Q4_1] = "q4_1",
[GGML_TYPE_Q4_2] = "q4_2",
[GGML_TYPE_Q4_3] = "q4_3",
[GGML_TYPE_Q5_0] = "q5_0",
[GGML_TYPE_Q5_1] = "q5_1",
[GGML_TYPE_Q8_0] = "q8_0",
@@ -3110,6 +3118,8 @@ static bool GGML_IS_QUANTIZED[GGML_TYPE_COUNT] = {
[GGML_TYPE_F16] = false,
[GGML_TYPE_Q4_0] = true,
[GGML_TYPE_Q4_1] = true,
[GGML_TYPE_Q4_2] = true,
[GGML_TYPE_Q4_3] = true,
[GGML_TYPE_Q5_0] = true,
[GGML_TYPE_Q5_1] = true,
[GGML_TYPE_Q8_0] = true,
@@ -6396,6 +6406,8 @@ static void ggml_compute_forward_add(
} break;
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q4_2:
case GGML_TYPE_Q4_3:
case GGML_TYPE_Q5_0:
case GGML_TYPE_Q5_1:
case GGML_TYPE_Q8_0:
@@ -7950,6 +7962,8 @@ static void ggml_compute_forward_mul_mat(
switch (src0->type) {
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q4_2:
case GGML_TYPE_Q4_3:
case GGML_TYPE_Q5_0:
case GGML_TYPE_Q5_1:
case GGML_TYPE_Q8_0:
@@ -8180,6 +8194,8 @@ static void ggml_compute_forward_get_rows(
switch (src0->type) {
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q4_2:
case GGML_TYPE_Q4_3:
case GGML_TYPE_Q5_0:
case GGML_TYPE_Q5_1:
case GGML_TYPE_Q8_0:
@@ -8504,6 +8520,8 @@ static void ggml_compute_forward_alibi(
} break;
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q4_2:
case GGML_TYPE_Q4_3:
case GGML_TYPE_Q5_0:
case GGML_TYPE_Q5_1:
case GGML_TYPE_Q8_0: