mikhail/llama.cpp-mtp-turboquant
1
0
Fork 0
Code Issues Pull Requests Actions 28 Packages Projects Releases Wiki Activity

CUDA & CPU: support F32 kernel type for CONV_TRANSPOSE_2D (#17094)

Browse Source 
* Refactor CUDA 2D transpose implementation to support multiple kernel types and improve parameter handling

- Introduced a `conv2d_transpose_params` struct for better parameter management.
- Updated `conv2d_transpose_kernel` to be templated for different kernel types (float and half).
- Modified `ggml_cuda_conv_2d_transpose_p0` to handle both F16 and F32 kernel types.
- Enhanced test cases to validate functionality for both kernel types.

* Refactor test cases for 2D convolution transpose to support dynamic kernel types

- Updated `test_conv_transpose_2d` structure to improve parameter handling by reordering constructor arguments.
- Enhanced test case generation to iterate over kernel types, allowing for flexible testing of different configurations.
- Removed hardcoded kernel type instances in favor of a loop for better maintainability and scalability.

* Refactor ggml_compute_forward_conv_transpose_2d to support both F16 and F32 tensor types.

* Refactor conv2d transpose kernel to use a template for kernel type, enhancing flexibility for different data types.
Update test cases to include both F16 and F32 tensor types for comprehensive coverage.

* Update ggml/src/ggml-cuda/conv2d-transpose.cu

Co-authored-by: Aman Gupta <amangupta052@gmail.com>

* Update ggml/src/ggml-cpu/ggml-cpu.c

Co-authored-by: Aman Gupta <amangupta052@gmail.com>

* Refactor conv2d transpose implementation by removing the conv2d_transpose_params struct and dispatching with direct kernel launch.

* Enhance cpu conv2d transpose implementation by introducing a templated kernel type for improved flexibility with F16 and F32 data types.

---------

Co-authored-by: Aman Gupta <amangupta052@gmail.com>
This commit is contained in:
Yihao Wang
2026-03-25 19:19:14 -07:00
committed by GitHub
parent c0159f9c1f
commit 0a524f2404
5 changed files with 122 additions and 53 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test-backend-ops.cpp
+21 -12
View File
@@ -4823,28 +4823,33 @@ struct test_conv_transpose_1d : public test_case {
// GGML_OP_CONV_TRANSPOSE_2D
struct test_conv_transpose_2d : public test_case {
// Dimensions
const std::array<int64_t, 4> ne_input;
const std::array<int64_t, 4> ne_kernel;
const int stride;
// Types
const ggml_type kernel_type;
std::string vars() override {
return VARS_TO_STR3(ne_input, ne_kernel, stride);
return VARS_TO_STR4(kernel_type, ne_input, ne_kernel, stride);
}
double max_nmse_err() override {
return 5e-4; // The default 1e-7 is too small for Vulkan.
}
test_conv_transpose_2d(std::array<int64_t, 4> ne_input = {10, 10, 3, 1}, // [input_width, input_height, input_channels, 1]
std::array<int64_t, 4> ne_kernel = {3, 3, 3, 1}, // [kernel_width, kernel_height, input_channels, 1]
int stride = 1)
: ne_input(ne_input), ne_kernel(ne_kernel), stride(stride){}
test_conv_transpose_2d(
std::array<int64_t, 4> ne_input = {10, 10, 3, 1}, // [input_width, input_height, input_channels, 1]
std::array<int64_t, 4> ne_kernel = {3, 3, 3, 1}, // [kernel_width, kernel_height, input_channels, 1]
int stride = 1,
ggml_type kernel_type = GGML_TYPE_F16
) : ne_input(ne_input), ne_kernel(ne_kernel), stride(stride), kernel_type(kernel_type) {}
ggml_tensor * build_graph(ggml_context * ctx) override {
ggml_tensor * input = ggml_new_tensor(ctx, GGML_TYPE_F32, 4, ne_input.data());
ggml_set_name(input, "input");
ggml_tensor * kernel = ggml_new_tensor(ctx, GGML_TYPE_F16, 4, ne_kernel.data());
ggml_tensor * kernel = ggml_new_tensor(ctx, kernel_type, 4, ne_kernel.data());
ggml_set_name(kernel, "kernel");
ggml_tensor * out = ggml_conv_transpose_2d_p0(ctx, kernel, input, stride);
@@ -7704,9 +7709,11 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
test_cases.emplace_back(new test_conv_transpose_1d({3,2,1,1}, {3,1,2,1}, 1, 0, 1));
test_cases.emplace_back(new test_conv_transpose_1d({2,1,1,1}, {3,1,1,1}, 1, 0, 1));
test_cases.emplace_back(new test_conv_transpose_2d({3, 2, 3, 1}, {2, 2, 1, 3}, 1));
test_cases.emplace_back(new test_conv_transpose_2d({10, 10, 9, 1}, {3, 3, 1, 9}, 2));
test_cases.emplace_back(new test_conv_transpose_2d({129, 63, 35, 1}, {3, 3, 48, 35}, 1));
for (ggml_type kernel_type : {GGML_TYPE_F32, GGML_TYPE_F16}) {
test_cases.emplace_back(new test_conv_transpose_2d({3, 2, 3, 1}, {2, 2, 1, 3}, 1, kernel_type));
test_cases.emplace_back(new test_conv_transpose_2d({10, 10, 9, 1}, {3, 3, 1, 9}, 2, kernel_type));
test_cases.emplace_back(new test_conv_transpose_2d({129, 63, 35, 1}, {3, 3, 48, 35}, 1, kernel_type));
}
test_cases.emplace_back(new test_count_equal(GGML_TYPE_F32, {4, 500, 1, 1}));
test_cases.emplace_back(new test_count_equal(GGML_TYPE_F32, {4, 5000, 1, 1}));
@@ -8892,9 +8899,11 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_perf() {
test_cases.emplace_back(new test_conv_2d_dw({512, 512, 256, 1}, {3, 3, 1, 256}, 1, 1, 1, false));
test_cases.emplace_back(new test_conv_2d_dw({512, 512, 256, 1}, {3, 3, 1, 256}, 1, 1, 1, true));
test_cases.emplace_back(new test_conv_transpose_2d({256, 256, 256, 1}, {3, 3, 16, 256}, 1));
test_cases.emplace_back(new test_conv_transpose_2d({16, 16, 16, 1}, {3, 3, 8, 16}, 1));
test_cases.emplace_back(new test_conv_transpose_2d({10, 10, 9, 1}, {3, 3, 1, 9}, 2));
for (ggml_type kernel_type : {GGML_TYPE_F32, GGML_TYPE_F16}) {
test_cases.emplace_back(new test_conv_transpose_2d({256, 256, 256, 1}, {3, 3, 16, 256}, 1, kernel_type));
test_cases.emplace_back(new test_conv_transpose_2d({16, 16, 16, 1}, {3, 3, 8, 16}, 1, kernel_type));
test_cases.emplace_back(new test_conv_transpose_2d({10, 10, 9, 1}, {3, 3, 1, 9}, 2, kernel_type));
}
test_cases.emplace_back(new test_mean(GGML_TYPE_F32, {256, 256, 3, 1}));