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

ggml : skip empty tensors in all backends

Browse source
This commit is contained in:
Francis Couture-Harpin 2024-03-18 21:12:53 -04:00
parent 4551e7eba8
commit 8b826c5b08
6 changed files with 16 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

2
ggml-cuda.cu
View file

@ -11354,7 +11354,7 @@ GGML_CALL static enum ggml_status ggml_backend_cuda_graph_compute(ggml_backend_t
for (int i = 0; i < cgraph->n_nodes; i++) { for (int i = 0; i < cgraph->n_nodes; i++) {
ggml_tensor * node = cgraph->nodes[i]; ggml_tensor * node = cgraph->nodes[i];
if (node->op == GGML_OP_RESHAPE || node->op == GGML_OP_TRANSPOSE || node->op == GGML_OP_VIEW || node->op == GGML_OP_PERMUTE || node->op == GGML_OP_NONE) { if (ggml_is_empty(node) || node->op == GGML_OP_RESHAPE || node->op == GGML_OP_TRANSPOSE || node->op == GGML_OP_VIEW || node->op == GGML_OP_PERMUTE || node->op == GGML_OP_NONE) {
continue; continue;
} }

4
ggml-kompute.cpp
View file

@ -1430,6 +1430,10 @@ static void ggml_vk_graph_compute(struct ggml_kompute_context * ctx, struct ggml
struct ggml_tensor * dst = gf->nodes[i]; struct ggml_tensor * dst = gf->nodes[i];
GGML_ASSERT(dst->data != nullptr); GGML_ASSERT(dst->data != nullptr);
if (ggml_is_empty(dst)) {
continue;
}
switch (dst->op) { switch (dst->op) {
case GGML_OP_NONE: case GGML_OP_NONE:
case GGML_OP_RESHAPE: case GGML_OP_RESHAPE:

4
ggml-metal.m
View file

@ -832,6 +832,10 @@ static enum ggml_status ggml_metal_graph_compute(
struct ggml_tensor * src2 = gf->nodes[i]->src[2]; struct ggml_tensor * src2 = gf->nodes[i]->src[2];
struct ggml_tensor * dst = gf->nodes[i]; struct ggml_tensor * dst = gf->nodes[i];
if (ggml_is_empty(dst)) {
continue;
}
switch (dst->op) { switch (dst->op) {
case GGML_OP_NONE: case GGML_OP_NONE:
case GGML_OP_RESHAPE: case GGML_OP_RESHAPE:

5
ggml-opencl.cpp
View file

@ -2234,6 +2234,11 @@ static ggml_backend_buffer_type_t ggml_backend_opencl_get_default_buffer_type(gg
static ggml_status ggml_backend_opencl_graph_compute(ggml_backend_t backend, ggml_cgraph * graph) { static ggml_status ggml_backend_opencl_graph_compute(ggml_backend_t backend, ggml_cgraph * graph) {
for (int i = 0; i < graph->n_nodes; ++i) { for (int i = 0; i < graph->n_nodes; ++i) {
ggml_tensor * node = graph->nodes[i]; ggml_tensor * node = graph->nodes[i];
if (ggml_is_empty(node)) {
continue;
}
switch (node->op) { switch (node->op) {
case GGML_OP_MUL_MAT: case GGML_OP_MUL_MAT:
ggml_cl_mul_mat(node->src[0], node->src[1], node, nullptr, 0); ggml_cl_mul_mat(node->src[0], node->src[1], node, nullptr, 0);

2
ggml-sycl.cpp
View file

@ -17229,7 +17229,7 @@ GGML_CALL static ggml_status ggml_backend_sycl_graph_compute(ggml_backend_t back
params.ith = 0; params.ith = 0;
for (int i = 0; i < cgraph->n_nodes; i++) { for (int i = 0; i < cgraph->n_nodes; i++) {
ggml_tensor * node = cgraph->nodes[i]; ggml_tensor * node = cgraph->nodes[i];
if (node->op == GGML_OP_RESHAPE || node->op == GGML_OP_TRANSPOSE || node->op == GGML_OP_VIEW || node->op == GGML_OP_PERMUTE || node->op == GGML_OP_NONE) { if (ggml_is_empty(node) || node->op == GGML_OP_RESHAPE || node->op == GGML_OP_TRANSPOSE || node->op == GGML_OP_VIEW || node->op == GGML_OP_PERMUTE || node->op == GGML_OP_NONE) {
continue; continue;
} }
#ifndef NDEBUG #ifndef NDEBUG

2
ggml-vulkan.cpp
View file

@ -5566,7 +5566,7 @@ GGML_CALL static ggml_status ggml_backend_vk_graph_compute(ggml_backend_t backen
for (int i = 0; i < cgraph->n_nodes; i++) { for (int i = 0; i < cgraph->n_nodes; i++) {
ggml_tensor * node = cgraph->nodes[i]; ggml_tensor * node = cgraph->nodes[i];
if (node->op == GGML_OP_RESHAPE || node->op == GGML_OP_TRANSPOSE || node->op == GGML_OP_VIEW || node->op == GGML_OP_PERMUTE || node->op == GGML_OP_NONE) { if (ggml_is_empty(node) || node->op == GGML_OP_RESHAPE || node->op == GGML_OP_TRANSPOSE || node->op == GGML_OP_VIEW || node->op == GGML_OP_PERMUTE || node->op == GGML_OP_NONE) {
continue; continue;
} }