ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-03-02 01:50:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

multi_add: simplify

Browse Source
This commit is contained in:
Iwan Kawrakow
2024-10-30 11:02:15 +02:00
parent 8af4111f97
commit c08af00548
4 changed files with 53 additions and 105 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
ggml/include/ggml.h
View File

@@ -933,7 +933,8 @@ extern "C" {
GGML_API struct ggml_tensor * ggml_multi_add(
struct ggml_context * ctx,
struct ggml_tensor ** a);
struct ggml_tensor * a,
int n_experts);
// dst = a
// view(dst, nb1, nb2, nb3, offset) += b

34
ggml/src/ggml-cuda/unary.cu
View File

@@ -62,9 +62,13 @@ static __global__ void multi_add_f32(int nused, int64_t ne0, int64_t ne1, int64_
int i0 = i % ne0;
float * result = (float *)(dst + i1*nb1);
const float * s = (const float *)(src0 + i1*nb01) + i0;
float sum = 0;
for (int j = 0; j < nused; ++j) sum += s[j*ne0];
result[i0] = sum;
if (nused == 1) {
result[i0] = s[0];
} else {
float sum = s[0] + s[ne0];
for (int j = 2; j < nused; ++j) sum += s[j*ne0];
result[i0] = sum;
}
}
static __global__ void fused_mul_relu_f32(const float * x, const float * y, float * dst, const int k) {
@@ -243,29 +247,9 @@ void ggml_cuda_op_multi_add(ggml_backend_cuda_context & ctx, ggml_tensor * dst)
GGML_ASSERT(dst->type == GGML_TYPE_F32);
GGML_ASSERT(dst->ne[2] == 1 && dst->ne[3] == 1);
GGML_ASSERT(dst->nb[0] == sizeof(float));
int nused = 0;
for (int i = 0; i < GGML_MAX_SRC; ++i) {
ggml_tensor * src = dst->src[i];
if (src) {
GGML_ASSERT(src->type == GGML_TYPE_F32);
GGML_ASSERT(ggml_are_same_shape(src, dst));
GGML_ASSERT(src->ne[2] == 1 && src->ne[3] == 1);
GGML_ASSERT(src->nb[0] == sizeof(float));
++nused;
} else {
break;
}
}
GGML_ASSERT(nused >= 2);
int nused = dst->op_params[0];
GGML_ASSERT(nused >= 1);
const char * src0 = (const char *)dst->src[0]->data;
const int64_t nb01 = dst->src[0]->ne[0]*sizeof(float);
for (int i = 1; i < nused; ++i) {
GGML_ASSERT(dst->src[i]->nb[1] == dst->src[0]->nb[1]);
const char * src = (const char *)dst->src[i]->data;
GGML_ASSERT(src == src0 + i*nb01);
GGML_ASSERT(dst->src[i]->nb[1] == dst->src[0]->nb[1]);
}
//printf("%s: nused = %d\n", __func__, nused);
cudaStream_t stream = ctx.stream();
multi_add_f32_cuda(nused, dst->ne[0], dst->ne[1], dst->nb[1], dst->src[0]->nb[1], src0, (char *)dst->data, stream);
}

64
ggml/src/ggml.c
View File

@@ -5112,41 +5112,21 @@ struct ggml_tensor * ggml_add_inplace(
struct ggml_tensor * ggml_multi_add(
struct ggml_context * ctx,
struct ggml_tensor ** a) {
struct ggml_tensor * a,
int n_experts) {
bool is_node = false;
struct ggml_tensor * a_used[GGML_MAX_SRC];
int n_used = 0;
for (int i = 0; i < GGML_MAX_SRC; ++i) {
if (a[i]) {
a_used[n_used++] = a[i];
}
}
if (n_used < 2) {
if (n_experts < 1) {
GGML_ABORT("fatal error");
}
if (n_used == 2) {
return ggml_add(ctx, a_used[0], a_used[1]);
}
for (int i = 1; i < n_used; ++i) {
if (!ggml_are_same_shape(a_used[i], a_used[0])) {
GGML_ABORT("fayal error");
}
}
struct ggml_tensor * result = ggml_dup_tensor(ctx, a_used[0]);
struct ggml_tensor * result = ggml_dup_tensor(ctx, a);
result->op = GGML_OP_MULTI_ADD;
result->grad = is_node ? ggml_dup_tensor(ctx, result) : NULL;
for (int i = 0; i < n_used; ++i) {
result->src[i] = a_used[i];
}
for (int i = n_used; i < GGML_MAX_SRC; ++i) {
result->src[i] = NULL;
}
result->src[0] = a;
result->op_params[0] = n_experts;
return result;
}
@@ -10474,13 +10454,15 @@ static void ggml_compute_forward_multi_add_f32(
const struct ggml_compute_params * params,
struct ggml_tensor * dst) {
struct ggml_tensor * src = dst->src[0];
GGML_ASSERT(dst->nb[0] == sizeof(float));
for (int i = 0; i < GGML_MAX_SRC; ++i) {
if (dst->src[i]) {
GGML_ASSERT(ggml_are_same_shape(dst->src[i], dst));
GGML_ASSERT(dst->src[i]->nb[0] == sizeof(float));
}
}
GGML_ASSERT(src->nb[0] == sizeof(float));
GGML_ASSERT(ggml_are_same_shape(src, dst));
GGML_ASSERT(dst->ne[2] == 1 && dst->ne[3] == 1);
const int n_add = dst->op_params[0];
GGML_ASSERT(n_add > 0);
const int ith = params->ith;
const int nth = params->nth;
@@ -10495,22 +10477,14 @@ static void ggml_compute_forward_multi_add_f32(
const int ir1 = MIN(ir0 + dr, nr);
int64_t ne0 = dst->ne[0];
int64_t ne1 = dst->ne[1];
int64_t ne2 = dst->ne[2];
for (int ir = ir0; ir < ir1; ++ir) {
// src1 is broadcastable across src0 and dst in i1, i2, i3
const int64_t i3 = ir/(ne2*ne1);
const int64_t i2 = (ir - i3*ne2*ne1)/ne1;
const int64_t i1 = (ir - i3*ne2*ne1 - i2*ne1);
for (int i1 = ir0; i1 < ir1; ++i1) {
float * dst_ptr = (float *) ((char *) dst->data + i3*dst->nb[3] + i2*dst->nb[2] + i1*dst->nb[1] );
float * dst_ptr = (float *) ((char *) dst->data + i1*dst->nb[1] );
const float * data = (const float *) ((const char *)src->data + i1*src->nb[1]);
memset(dst_ptr, 0, ne0*sizeof(float));
for (int i = 0; i < GGML_MAX_SRC; ++i) {
struct ggml_tensor * src = dst->src[i];
if (!src) continue;
const float * data = (const float *) ((const char *) src->data + i3*src->nb[3] + i2*src->nb[2] + i1*src->nb[1]);
ggml_vec_add_f32(ne0, dst_ptr, dst_ptr, data);
for (int j = 0; j < n_add; ++j) {
ggml_vec_add_f32(ne0, dst_ptr, dst_ptr, data + j*ne0);
}
}
}

57
src/llama.cpp
View File

@@ -8358,44 +8358,33 @@ static struct ggml_tensor * llm_build_moe_ffn(
return ggml_add(ctx, ggml_view_2d(ctx, experts, n_embd, n_tokens, experts->nb[2], 0),
ggml_view_2d(ctx, experts, n_embd, n_tokens, experts->nb[2], experts->nb[1]));
}
if (n_expert_used <= GGML_MAX_SRC) {
ggml_tensor * src[GGML_MAX_SRC];
for (int i = 0; i < n_expert_used; ++i) {
src[i] = ggml_view_2d(ctx, experts, n_embd, n_tokens, experts->nb[2], i*experts->nb[1]);
}
for (int i = n_expert_used; i < GGML_MAX_SRC; ++i) src[i] = nullptr;
return ggml_multi_add(ctx, src);
}
return ggml_multi_add(ctx, ggml_view_2d(ctx, experts, n_embd, n_tokens, experts->nb[2], 0), n_expert_used);
GGML_ABORT("fatal error");
//// aggregate experts
//ggml_tensor * moe_out = nullptr;
////ggml_tensor * first_expert = nullptr;
//for (int i = 0; i < n_expert_used; ++i) {
// ggml_tensor * cur_expert = ggml_view_2d(ctx, experts, n_embd, n_tokens,
// experts->nb[2], i*experts->nb[1]);
//int nloop = (n_expert_used + GGML_MAX_SRC - 1)/GGML_MAX_SRC;
// if (i == 0) {
// moe_out = cur_expert;
// //first_expert = cur_expert;
// //printf("%s: %d: %d x %d x %d x %d | %d x %d x %d x %d\n", __func__, ggml_is_contiguous(first_expert),
// // (int)cur_expert->ne[0], (int)cur_expert->ne[1], (int)cur_expert->ne[2], (int)cur_expert->ne[3],
// // (int)cur_expert->nb[0], (int)cur_expert->nb[1], (int)cur_expert->nb[2], (int)cur_expert->nb[3]);
// } else {
// moe_out = ggml_add(ctx, moe_out, cur_expert);
// //printf("%s: %d %d\n", __func__, ggml_is_contiguous(cur_expert), ggml_are_same_shape(cur_expert, first_expert));
// }
//}
// aggregate experts
ggml_tensor * moe_out = nullptr;
//ggml_tensor * first_expert = nullptr;
for (int i = 0; i < n_expert_used; ++i) {
ggml_tensor * cur_expert = ggml_view_2d(ctx, experts, n_embd, n_tokens,
experts->nb[2], i*experts->nb[1]);
//if (n_expert_used == 1) {
// // avoid returning a non-contiguous tensor
// moe_out = ggml_cont(ctx, moe_out);
//}
if (i == 0) {
moe_out = cur_expert;
//first_expert = cur_expert;
//printf("%s: %d: %d x %d x %d x %d | %d x %d x %d x %d\n", __func__, ggml_is_contiguous(first_expert),
// (int)cur_expert->ne[0], (int)cur_expert->ne[1], (int)cur_expert->ne[2], (int)cur_expert->ne[3],
// (int)cur_expert->nb[0], (int)cur_expert->nb[1], (int)cur_expert->nb[2], (int)cur_expert->nb[3]);
} else {
moe_out = ggml_add(ctx, moe_out, cur_expert);
//printf("%s: %d %d\n", __func__, ggml_is_contiguous(cur_expert), ggml_are_same_shape(cur_expert, first_expert));
}
}
if (n_expert_used == 1) {
// avoid returning a non-contiguous tensor
moe_out = ggml_cont(ctx, moe_out);
}
return moe_out;
//return moe_out;
}
static struct ggml_tensor * llm_build_kqv(