mirror of
https://github.com/ggerganov/llama.cpp.git
synced 2025-01-05 00:04:36 +00:00
Minor cleanup.
This commit is contained in:
parent
601905e75e
commit
5509f74318
@ -939,7 +939,6 @@ void ggml_vk_mul_mat_q4_x(const std::vector<uint32_t>& spirv, uint32_t block_siz
|
|||||||
s_algo->setPushConstants<PushConstants>({pushConsts});
|
s_algo->setPushConstants<PushConstants>({pushConsts});
|
||||||
s_algo->updateDescriptors(s_kompute_context->pool.get());
|
s_algo->updateDescriptors(s_kompute_context->pool.get());
|
||||||
}
|
}
|
||||||
seq.record<kp::OpTensorFill>({out});
|
|
||||||
seq.record<kp::OpAlgoDispatch>(s_algo);
|
seq.record<kp::OpAlgoDispatch>(s_algo);
|
||||||
}
|
}
|
||||||
|
|
||||||
@ -951,7 +950,6 @@ void ggml_vk_mul_mat_q4_0(Args&&... args) {
|
|||||||
ggml_vk_mul_mat_q4_x(spirv, 1/*We access blocks unaligned*/, std::forward<Args>(args)...);
|
ggml_vk_mul_mat_q4_x(spirv, 1/*We access blocks unaligned*/, std::forward<Args>(args)...);
|
||||||
}
|
}
|
||||||
|
|
||||||
// FIXME: This could be improved like was done in q4_0 version but needs testing...
|
|
||||||
template <typename... Args>
|
template <typename... Args>
|
||||||
void ggml_vk_mul_mat_q4_1(Args&&... args) {
|
void ggml_vk_mul_mat_q4_1(Args&&... args) {
|
||||||
const static auto spirv = getSpirvShader(kp::shader_data::op_mul_mat_q4_1_comp_spv,
|
const static auto spirv = getSpirvShader(kp::shader_data::op_mul_mat_q4_1_comp_spv,
|
||||||
|
Loading…
Reference in New Issue
Block a user