vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

ggml : mul_mat_id use the same tensor for all the experts (#6387)

Browse source 
* ggml : update mul_mat_id to use the same tensor for all the experts

* update cuda

* minor

* update metal

* update test-backend-ops

* fix cuda

* Update ggml-metal.m

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

* update convert.py

* update convert-hf-to-gguf.py

* update convert.py for mixtral hf models

* Update convert-hf-to-gguf.py

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

* cuda : support non-pow-2 number of experts

* allow quantize to work for split and merged experts models in the same way

* cleanup + disable mmap automatically with split tensors models

* update imatrix

* test-backend-ops : test qwen argsort

* update grok model loading

* llama : add merged experts tensors to the grok tensor map

* minor

* gguf : bump version

* fix quantizing of merged experts

* convert-hf-to-gguf.py : update grok (untested)

* make linter happy

* cuda/argsort : use shared memory instead of pool memory

* convert : fix grok tensor names

* metal : add support for non-pow-2 argsort

* llama : more loader cleanup, better error checking

* cuda : fix warning

* llama : still use mmap for loading old models, but copy the data to a host buffer

* add review note

* llama : remove ffn tensor counting + add sanity check

ggml-ci

* convert : fix handling of n_experts == None

ggml-ci

* imatrix : fix ncall counters

* llama : produce error if imatrix size does not match

* quantize : terminate on errors + trace logs

ggml-ci

* metal : pad shared memory to 16 bytes

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
slaren 2024-04-03 15:07:05 +02:00 • committed by GitHub
parent 52604860f9
commit 08a0c02060
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
15 changed files with 756 additions and 888 deletions
Download patch file Download diff file Expand all files Collapse all files

52
ggml-cuda/argsort.cu
View file

@ -8,32 +8,41 @@ static inline __device__ void ggml_cuda_swap(T & a, T & b) {
}
template<ggml_sort_order order>
static __global__ void k_argsort_f32_i32(const float * x, int * dst, const int ncols) {
static __global__ void k_argsort_f32_i32(const float * x, int * dst, const int ncols, int ncols_pad) {
// bitonic sort
int col = threadIdx.x;
int row = blockIdx.y;
if (col >= ncols) return;
if (col >= ncols_pad) {
return;
}
const float * x_row = x + row * ncols;
int * dst_row = dst + row * ncols;
extern __shared__ int dst_row[];
// initialize indices
if (col < ncols) {
dst_row[col] = col;
}
dst_row[col] = col;
__syncthreads();
for (int k = 2; k <= ncols; k *= 2) {
for (int k = 2; k <= ncols_pad; k *= 2) {
for (int j = k / 2; j > 0; j /= 2) {
int ixj = col ^ j;
if (ixj > col) {
if ((col & k) == 0) {
if (order == GGML_SORT_ORDER_ASC ? x_row[dst_row[col]] > x_row[dst_row[ixj]] : x_row[dst_row[col]] < x_row[dst_row[ixj]]) {
if (dst_row[col] >= ncols ||
(dst_row[ixj] < ncols && (order == GGML_SORT_ORDER_ASC ?
x_row[dst_row[col]] > x_row[dst_row[ixj]] :
x_row[dst_row[col]] < x_row[dst_row[ixj]]))
) {
ggml_cuda_swap(dst_row[col], dst_row[ixj]);
}
} else {
if (order == GGML_SORT_ORDER_ASC ? x_row[dst_row[col]] < x_row[dst_row[ixj]] : x_row[dst_row[col]] > x_row[dst_row[ixj]]) {
if (dst_row[ixj] >= ncols ||
(dst_row[col] < ncols && (order == GGML_SORT_ORDER_ASC ?
x_row[dst_row[col]] < x_row[dst_row[ixj]] :
x_row[dst_row[col]] > x_row[dst_row[ixj]]))
) {
ggml_cuda_swap(dst_row[col], dst_row[ixj]);
}
}
@ -41,18 +50,35 @@ static __global__ void k_argsort_f32_i32(const float * x, int * dst, const int n
__syncthreads();
}
}
// copy the result to dst without the padding
if (col < ncols) {
dst[row * ncols + col] = dst_row[col];
}
}
static int next_power_of_2(int x) {
int n = 1;
while (n < x) {
n *= 2;
}
return n;
}
static void argsort_f32_i32_cuda(const float * x, int * dst, const int ncols, const int nrows, ggml_sort_order order, cudaStream_t stream) {
// bitonic sort requires ncols to be power of 2
GGML_ASSERT((ncols & (ncols - 1)) == 0);
const int ncols_pad = next_power_of_2(ncols);
const dim3 block_dims(ncols, 1, 1);
const dim3 block_dims(ncols_pad, 1, 1);
const dim3 block_nums(1, nrows, 1);
const size_t shared_mem = ncols_pad * sizeof(int);
GGML_ASSERT(shared_mem <= ggml_cuda_info().devices[ggml_cuda_get_device()].smpb);
if (order == GGML_SORT_ORDER_ASC) {
k_argsort_f32_i32<GGML_SORT_ORDER_ASC><<<block_nums, block_dims, 0, stream>>>(x, dst, ncols);
k_argsort_f32_i32<GGML_SORT_ORDER_ASC><<<block_nums, block_dims, shared_mem, stream>>>(x, dst, ncols, ncols_pad);
} else if (order == GGML_SORT_ORDER_DESC) {
k_argsort_f32_i32<GGML_SORT_ORDER_DESC><<<block_nums, block_dims, 0, stream>>>(x, dst, ncols);
k_argsort_f32_i32<GGML_SORT_ORDER_DESC><<<block_nums, block_dims, shared_mem, stream>>>(x, dst, ncols, ncols_pad);
} else {
GGML_ASSERT(false);
}