Public/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-09-18 16:55:05 +02:00
Code Issues Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'upstream/master' into backend-sampling

Browse Source
This commit is contained in:
Daniel Bevenius
2025-11-20 06:57:24 +01:00
parent 18ed4d8f96 7d77f07325
commit 0c660e7390
64 changed files with 5906 additions and 1409 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test-backend-ops.cpp
+38 -18
View File
@@ -2776,24 +2776,34 @@ struct test_cpy : public test_case {
struct test_cont : public test_case {
const ggml_type type;
const std::array<int64_t, 4> ne;
bool use_view_slice;
std::string vars() override {
return VARS_TO_STR2(type, ne);
return VARS_TO_STR3(type, ne, use_view_slice);
}
test_cont(ggml_type type = GGML_TYPE_F32,
std::array<int64_t, 4> ne = {10, 10, 10, 1})
: type(type), ne(ne) {}
std::array<int64_t, 4> ne = {10, 10, 10, 1},
bool use_view_slice = false)
: type(type), ne(ne), use_view_slice(use_view_slice) {}
ggml_tensor * build_graph(ggml_context * ctx) override {
ggml_tensor * src = ggml_new_tensor(ctx, type, 4, ne.data());
ggml_set_param(src);
ggml_set_name(src, "src");
src = ggml_transpose(ctx, src);
ggml_set_name(src, "src_transposed");
ggml_tensor * out = ggml_cont(ctx, src);
ggml_tensor * dst;
if (use_view_slice) {
dst = ggml_view_4d(ctx, src, src->ne[0], 1, src->ne[2], src->ne[3],
src->nb[1], src->nb[2], src->nb[3], src->nb[0] * (src->ne[1] - 1));
ggml_set_name(dst, "src_view_slice");
} else {
dst = ggml_transpose(ctx, src);
ggml_set_name(dst, "src_transposed");
}
ggml_tensor * out = ggml_cont(ctx, dst);
ggml_set_name(out, "out");
return out;
@@ -6945,16 +6955,17 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
test_cases.emplace_back(new test_cpy(GGML_TYPE_BF16, GGML_TYPE_BF16, {256, 4, 1, 1}, {0, 0, 0, 0}, {0, 0, 0, 0}, true));
test_cases.emplace_back(new test_cpy(GGML_TYPE_F32, GGML_TYPE_F32, {256, 1, 4, 1}, {1, 2, 0, 3}, {0, 0, 0, 0}));
test_cases.emplace_back(new test_cont());
test_cases.emplace_back(new test_cont(GGML_TYPE_F32, {2, 1, 1 ,1}));
test_cases.emplace_back(new test_cont(GGML_TYPE_F32, {2, 1, 3 ,5}));
test_cases.emplace_back(new test_cont(GGML_TYPE_F32, {2, 3, 5 ,7}));
test_cases.emplace_back(new test_cont(GGML_TYPE_F16, {2, 1, 1 ,1}));
test_cases.emplace_back(new test_cont(GGML_TYPE_F16, {2, 1, 3 ,5}));
test_cases.emplace_back(new test_cont(GGML_TYPE_F16, {2, 3, 5 ,7}));
test_cases.emplace_back(new test_cont(GGML_TYPE_BF16, {2, 1, 1 ,1}));
test_cases.emplace_back(new test_cont(GGML_TYPE_BF16, {2, 1, 3 ,5}));
test_cases.emplace_back(new test_cont(GGML_TYPE_BF16, {2, 3, 5 ,7}));
for (ggml_type type_dst : { GGML_TYPE_F32, GGML_TYPE_F16, GGML_TYPE_BF16 }) {
for (bool use_view_slice : { true, false }) {
for (std::array<int64_t, 4> ne : std::initializer_list<std::array<int64_t, 4>>{ {2, 1, 1, 1}, {2, 1, 3, 5},
{2, 3, 5, 7}, {1, 4, 4, 1}, {1, 8, 17, 1}, {10, 10, 10, 1} }) {
if (use_view_slice && (type_dst == GGML_TYPE_F16 || type_dst == GGML_TYPE_BF16)) {
continue; // TODO: add after WebGPU is fixed
}
test_cases.emplace_back(new test_cont(type_dst, ne, use_view_slice));
}
}
}
auto add_test_bin_bcast = [&](ggml_type type, std::array<int64_t, 4> ne, std::array<int, 4> nr) {
for (auto op : {ggml_add, ggml_sub, ggml_mul, ggml_div}) {
@@ -7015,6 +7026,7 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
test_cases.emplace_back(new test_bin_bcast(ggml_add, GGML_TYPE_F32, {16, 5, 4, 3}, {1, 1, 1, 1}, 16));
test_cases.emplace_back(new test_add1());
test_cases.emplace_back(new test_add1(GGML_TYPE_F32, {1024, 1024, 1, 1}));
test_cases.emplace_back(new test_scale());
test_cases.emplace_back(new test_scale(GGML_TYPE_F32, {10, 10, 10, 10}, 2.0f, 1.0f));
test_cases.emplace_back(new test_scale(GGML_TYPE_F32, {10, 10, 10, 10}, 2.0f, 1.0f, true)); // inplace test
@@ -7354,9 +7366,13 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
test_cases.emplace_back(new test_clamp (type, {7, 1, 5, 3}));
test_cases.emplace_back(new test_leaky_relu(type, {7, 1, 5, 3}));
test_cases.emplace_back(new test_floor (type, {7, 1, 5, 3}));
test_cases.emplace_back(new test_floor (type, { 1024, 1024, 1, 1 }));
test_cases.emplace_back(new test_ceil (type, {7, 1, 5, 3}));
test_cases.emplace_back(new test_ceil (type, { 1024, 1024, 1, 1 }));
test_cases.emplace_back(new test_round (type, {7, 1, 5, 3}));
test_cases.emplace_back(new test_round (type, { 1024, 1024, 1, 1 }));
test_cases.emplace_back(new test_trunc (type, {7, 1, 5, 3}));
test_cases.emplace_back(new test_trunc (type, { 1024, 1024, 1, 1 }));
}
test_cases.emplace_back(new test_diag_mask_inf(GGML_TYPE_F32, {10, 10, 1, 1}, 5));
@@ -7501,13 +7517,15 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
}
for (ggml_sort_order order : {GGML_SORT_ORDER_ASC, GGML_SORT_ORDER_DESC}) {
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {8, 1, 1, 1}, order));
for (uint32_t i = 4; i <= 1024*1024; i *= 2) {
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {i-1, 1, 1, 1}));
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {i, 1, 1, 1}));
}
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {16, 10, 10, 10}, order));
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {60, 10, 10, 10}, order)); // qwen
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {1023, 2, 1, 3}, order));
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {1024, 2, 1, 3}, order));
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {1025, 2, 1, 3}, order));
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {16384, 1, 1, 1}, order)); // many backends only handle up to 1024
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {2047, 2, 1, 3}, order));
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {2048, 2, 1, 3}, order));
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {2049, 2, 1, 3}, order));
@@ -7556,6 +7574,7 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
test_cases.emplace_back(new test_pad_reflect_1d(GGML_TYPE_F32, {3000, 384, 4, 1}));
test_cases.emplace_back(new test_roll());
test_cases.emplace_back(new test_arange());
test_cases.emplace_back(new test_arange(GGML_TYPE_F32, 0.0f, 1048576.0f, 1.0f));
test_cases.emplace_back(new test_timestep_embedding());
test_cases.emplace_back(new test_leaky_relu());
@@ -7583,6 +7602,7 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
test_cases.emplace_back(new test_fill(0.0f));
test_cases.emplace_back(new test_fill(2.0f, GGML_TYPE_F32, { 303, 207, 11, 3 }));
test_cases.emplace_back(new test_fill(-152.0f, GGML_TYPE_F32, { 800, 600, 4, 4 }));
test_cases.emplace_back(new test_fill(3.5f, GGML_TYPE_F32, { 2048, 512, 2, 2 }));
test_cases.emplace_back(new test_solve_tri());
test_cases.emplace_back(new test_solve_tri(GGML_TYPE_F32, { 11, 11, 1, 1 }, { 5, 11, 1, 1 }));