mirror of
https://github.com/ggerganov/llama.cpp.git
synced 2024-11-15 07:19:53 +00:00
f3f65429c4
* scripts : update sync [no ci] * files : relocate [no ci] * ci : disable kompute build [no ci] * cmake : fixes [no ci] * server : fix mingw build ggml-ci * cmake : minor [no ci] * cmake : link math library [no ci] * cmake : build normal ggml library (not object library) [no ci] * cmake : fix kompute build ggml-ci * make,cmake : fix LLAMA_CUDA + replace GGML_CDEF_PRIVATE ggml-ci * move public backend headers to the public include directory (#8122) * move public backend headers to the public include directory * nix test * spm : fix metal header --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * scripts : fix sync paths [no ci] * scripts : sync ggml-blas.h [no ci] --------- Co-authored-by: slaren <slarengh@gmail.com>
40 lines
962 B
Plaintext
40 lines
962 B
Plaintext
#version 450
|
|
|
|
#include "common.comp"
|
|
|
|
#define NL 2
|
|
#define BYTES_FOR_TYPE 4 /*bytes for float*/
|
|
#define SIZE_OF_BLOCK sizeof_block_q4_1
|
|
|
|
layout(local_size_x = 1) in;
|
|
|
|
layout (binding = 0) readonly buffer tensorInA { uint8_t inA[]; };
|
|
layout (binding = 1) readonly buffer tensorInB { int inB[]; };
|
|
layout (binding = 2) writeonly buffer tensorOut { float out_[]; };
|
|
|
|
layout (push_constant) uniform parameter {
|
|
uint inAOff;
|
|
uint inBOff;
|
|
uint outOff;
|
|
int ne00;
|
|
int nb01;
|
|
int nb1;
|
|
} pcs;
|
|
|
|
block_q4_1 get_unaligned_block_q4_1(uint index) {
|
|
block_q4_1 fres;
|
|
fres.d = u8BufToFloat16(inA, index);
|
|
fres.m = u8BufToFloat16(inA, index+2);
|
|
[[unroll]] for (uint it = 0; it != QK4_1 / 2; it++) {
|
|
fres.qs[it] = inA[index+4+it];
|
|
}
|
|
return fres;
|
|
}
|
|
|
|
mat4 dequantize_block(uint index, uint il) {
|
|
const block_q4_1 block = get_unaligned_block_q4_1(index);
|
|
return dequantize_q4_1(block, il);
|
|
}
|
|
|
|
#include "op_getrows.comp"
|