Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Add Q2_0 quantization: type definition and CPU backend (#24448)

Browse Source
This commit is contained in:
Pasha Khosravi
2026-07-07 12:05:47 -07:00
committed by GitHub
parent c198af4dc2
commit bec4772f6a
17 changed files with 262 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-cpu/ops.cpp
+7
View File
@@ -665,6 +665,7 @@ void ggml_compute_forward_add(
ggml_compute_forward_add_non_quantized(params, dst);
} break;
case GGML_TYPE_Q1_0:
case GGML_TYPE_Q2_0:
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q5_0:
@@ -1115,6 +1116,7 @@ void ggml_compute_forward_add1(
}
} break;
case GGML_TYPE_Q1_0:
case GGML_TYPE_Q2_0:
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q5_0:
@@ -1245,6 +1247,7 @@ void ggml_compute_forward_acc(
case GGML_TYPE_F16:
case GGML_TYPE_BF16:
case GGML_TYPE_Q1_0:
case GGML_TYPE_Q2_0:
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q5_0:
@@ -4454,6 +4457,7 @@ void ggml_compute_forward_out_prod(
switch (src0->type) {
case GGML_TYPE_Q1_0:
case GGML_TYPE_Q2_0:
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q5_0:
@@ -4730,6 +4734,7 @@ void ggml_compute_forward_set(
case GGML_TYPE_F16:
case GGML_TYPE_BF16:
case GGML_TYPE_Q1_0:
case GGML_TYPE_Q2_0:
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q5_0:
@@ -4954,6 +4959,7 @@ void ggml_compute_forward_get_rows(
switch (src0->type) {
case GGML_TYPE_Q1_0:
case GGML_TYPE_Q2_0:
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q5_0:
@@ -5680,6 +5686,7 @@ void ggml_compute_forward_clamp(
} break;
case GGML_TYPE_BF16:
case GGML_TYPE_Q1_0:
case GGML_TYPE_Q2_0:
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q5_0: