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

ggml : add ggml_upscale_ext (ggml/814)

Browse source 
* initial commit with CPU implementation of upscale to shape and test, cuda implementation next

* experimental commit to see if dst shape is correct

* test version

* test

* removed unnecessary params

* refactor

* fixed tests

* ggml : metal impl + cleanup + sycl dev warnings

* patched ggml_upscale cuda op to handle non-contiguous tensors, added test for non-contiguous behavior

* metal : fix upsacle op to support nb00 + style

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
John Balis 2024-05-15 03:52:33 -05:00 • committed by Georgi Gerganov
parent 583fd6b000
commit 48aa8fd1f2
No known key found for this signature in database
GPG key ID: BF970631944C16B7
7 changed files with 146 additions and 60 deletions
Download patch file Download diff file Expand all files Collapse all files

32
tests/test-backend-ops.cpp
View file

@ -1329,23 +1329,47 @@ struct test_upscale : public test_case {
const ggml_type type;
const std::array<int64_t, 4> ne;
const int32_t scale_factor;
const bool transpose;
std::string vars() override {
return VARS_TO_STR3(type, ne, scale_factor);
return VARS_TO_STR4(type, ne, scale_factor, transpose);
}
test_upscale(ggml_type type = GGML_TYPE_F32,
std::array<int64_t, 4> ne = {512, 512, 3, 1},
int32_t scale_factor = 2)
: type(type), ne(ne), scale_factor(scale_factor) {}
int32_t scale_factor = 2, bool transpose = false)
: type(type), ne(ne), scale_factor(scale_factor), transpose(transpose) {}
ggml_tensor * build_graph(ggml_context * ctx) override {
ggml_tensor * a = ggml_new_tensor(ctx, type, 4, ne.data());
if (transpose) a = ggml_transpose(ctx, a);
ggml_tensor * out = ggml_upscale(ctx, a, scale_factor);
return out;
}
};
// GGML_OP_UPSCALE (ext)
struct test_upscale_ext : public test_case {
const ggml_type type;
const std::array<int64_t, 4> ne;
const std::array<int64_t, 4> ne_tgt;
std::string vars() override {
return VARS_TO_STR3(type, ne, ne_tgt);
}
test_upscale_ext(ggml_type type = GGML_TYPE_F32,
std::array<int64_t, 4> ne = {2, 5, 7, 11},
std::array<int64_t, 4> ne_tgt = {5, 7, 11, 13})
: type(type), ne(ne), ne_tgt(ne_tgt) {}
ggml_tensor * build_graph(ggml_context * ctx) override {
ggml_tensor * a = ggml_new_tensor(ctx, type, 4, ne.data());
ggml_tensor * out = ggml_upscale_ext(ctx, a, ne_tgt[0], ne_tgt[1],ne_tgt[2], ne_tgt[3]);
return out;
}
};
// GGML_OP_GROUP_NORM
struct test_group_norm : public test_case {
const ggml_type type;
@ -2169,6 +2193,8 @@ static bool test_backend(ggml_backend_t backend, test_mode mode, const char * op
test_cases.emplace_back(new test_sum_rows());
test_cases.emplace_back(new test_upscale());
test_cases.emplace_back(new test_upscale(GGML_TYPE_F32, { 512, 512, 3, 1 }, 2, true));
test_cases.emplace_back(new test_upscale_ext());
test_cases.emplace_back(new test_group_norm());
test_cases.emplace_back(new test_acc());
test_cases.emplace_back(new test_pad());