diff options
author | Kawrakow <iwankawrakow@gmail.com> | 2025-06-05 08:31:20 +0300 |
---|---|---|
committer | GitHub <noreply@github.com> | 2025-06-05 08:31:20 +0300 |
commit | 8ffad187abbb93b74db8ef813b6fdceec80e02b0 (patch) | |
tree | 2c078dfcbe2dd36b46675651ae5d91abd41641f4 /ggml/src/ggml-cuda/common.cuh | |
parent | 0b10f7418f7315ef90e35da49e0c053b395fd528 (diff) |
MMQ implementation for IQ4_KS_R4 and IQ5_KS_R4 (#493)
* MMQ for iq4_ks_r4
* MMQ for iq5_ks_r4
* Add forgotten file
* Another forgotten file
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Diffstat (limited to 'ggml/src/ggml-cuda/common.cuh')
-rw-r--r-- | ggml/src/ggml-cuda/common.cuh | 14 |
1 files changed, 14 insertions, 0 deletions
diff --git a/ggml/src/ggml-cuda/common.cuh b/ggml/src/ggml-cuda/common.cuh index bc7fadb0..8f3d2a26 100644 --- a/ggml/src/ggml-cuda/common.cuh +++ b/ggml/src/ggml-cuda/common.cuh @@ -600,6 +600,13 @@ struct ggml_cuda_type_traits<GGML_TYPE_IQ4_KS> { }; template<> +struct ggml_cuda_type_traits<GGML_TYPE_IQ4_KS_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_KSS> { static constexpr int qk = QK_K; static constexpr int qr = QR4_XS; @@ -621,6 +628,13 @@ struct ggml_cuda_type_traits<GGML_TYPE_IQ5_KS> { }; template<> +struct ggml_cuda_type_traits<GGML_TYPE_IQ5_KS_R4> { + static constexpr int qk = QK_K; + static constexpr int qr = QR5_XS; + static constexpr int qi = QI5_XS; +}; + +template<> struct ggml_cuda_type_traits<GGML_TYPE_IQ6_K> { static constexpr int qk = QK_K; static constexpr int qr = QR6_XS; |