summaryrefslogtreecommitdiff
path: root/ggml/src/ggml-cuda/common.cuh
diff options
context:
space:
mode:
Diffstat (limited to 'ggml/src/ggml-cuda/common.cuh')
-rw-r--r--ggml/src/ggml-cuda/common.cuh29
1 files changed, 29 insertions, 0 deletions
diff --git a/ggml/src/ggml-cuda/common.cuh b/ggml/src/ggml-cuda/common.cuh
index a0cdab28..e2690cb3 100644
--- a/ggml/src/ggml-cuda/common.cuh
+++ b/ggml/src/ggml-cuda/common.cuh
@@ -662,6 +662,35 @@ struct ggml_cuda_type_traits<GGML_TYPE_IQ3_S> {
static constexpr int qi = QI3_S;
};
+template<>
+struct ggml_cuda_type_traits<GGML_TYPE_IQ2_K_R4> {
+ static constexpr int qk = QK_K;
+ static constexpr int qr = QR4_XS;
+ static constexpr int qi = QI4_XS;
+};
+
+template<>
+struct ggml_cuda_type_traits<GGML_TYPE_IQ3_K_R4> {
+ static constexpr int qk = QK_K;
+ static constexpr int qr = QR4_XS;
+ static constexpr int qi = QI4_XS;
+};
+
+template<>
+struct ggml_cuda_type_traits<GGML_TYPE_IQ4_K_R4> {
+ static constexpr int qk = QK_K;
+ static constexpr int qr = QR4_XS;
+ static constexpr int qi = QI4_XS;
+};
+
+template<>
+struct ggml_cuda_type_traits<GGML_TYPE_IQ5_K_R4> {
+ static constexpr int qk = QK_K;
+ static constexpr int qr = QR5_XS;
+ static constexpr int qi = QI5_XS;
+};
+
+
//////////////////////
struct ggml_cuda_device_info {