Skip to content

build: repair architecture tests and package configuration #80

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
lalalune merged 2 commits into from
Oct 14, 2026
Merged

build: repair architecture tests and package configuration #80

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
07852b6
tests: use current NextN embedding API in architecture fixture
lalalune Oct 14, 2026
1aaf50f
build: repair package config nesting and Kokoro rope classification
lalalune Oct 14, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion ggml/cmake/ggml-config.cmake.in
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -188,7 +188,7 @@ if(NOT TARGET ggml::ggml)
if (TARGET ggml::${_ggml_backend})
list(APPEND _ggml_all_targets ggml::${_ggml_backend})
endif()
endforeach()
endif()

# Synthetic ggml::all aggregating every available backend — matches the
# upstream contract used by examples/simple-cmake-pkg and any external
Expand Down
1 change: 1 addition & 0 deletions src/llama-model.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2428,6 +2428,7 @@ llama_rope_type llama_model_rope_type(const llama_model * model) {
switch (model->arch) {
// these models do not use RoPE
case LLM_ARCH_CLIP:
case LLM_ARCH_KOKORO:
case LLM_ARCH_GPT2:
case LLM_ARCH_GPTJ:
case LLM_ARCH_MPT:
Expand Down
6 changes: 3 additions & 3 deletions tests/test-llama-archs.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -341,7 +341,7 @@ static std::vector<float> get_logits(
}
if (target_ctx) {
llama_memory_clear(llama_get_memory(target_source), true);
llama_set_embeddings_pre_norm(target_source, true);
llama_set_embeddings_nextn(target_source, true, false);
get_logits(llama_get_model(target_source), target_source, tokens);
if (target_source != target_ctx) {
std::vector<uint8_t> state(llama_state_get_size(target_source));
Expand All @@ -366,10 +366,10 @@ static std::vector<float> get_logits(
batch.token = static_cast<llama_token *>(std::malloc(n_ctx * sizeof(llama_token)));
GGML_ASSERT(batch.token);
// Match MTP prefill: first row has no prior state; later rows consume
// the real target's preceding pre-normalization hidden state.
// the real target's preceding NextN hidden state.
std::memset(batch.embd, 0, n_embd * sizeof(float));
for (uint32_t i = 1; i < n_tokens; ++i) {
const float * hidden = llama_get_embeddings_pre_norm_ith(target_source, i - 1);
const float * hidden = llama_get_embeddings_nextn_ith(target_source, i - 1);
GGML_ASSERT(hidden);
std::memcpy(batch.embd + i * n_embd, hidden, n_embd * sizeof(float));
}
Expand Down
Loading