Skip to content

Enable minicpm3 for OpenVINO backend architecture tests #175

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
zhaixuejun1993 merged 1 commit into from
May 20, 2026
Merged

Enable minicpm3 for OpenVINO backend architecture tests #175

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
12 changes: 10 additions & 2 deletions ggml/src/ggml-openvino/ggml-decoder.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -463,6 +463,10 @@ std::pair<ModelParams, ComputeParams> GgmlOvDecoder::compute_llm_params(ggml_cgr
}
}
if (node->op == GGML_OP_ROPE) {
if (compute_params.token_len_per_seq == -1 && node->src[1] != nullptr) {
compute_params.token_len_per_seq = ggml_nelements(node->src[1]);
}

// When multiple ROPE ops in the graph disagree on op_params (e.g. gemma4's
// mixed SWA/non-SWA layers with different n_dims or freq_base), we cannot
// share a single precomputed rope_sin/rope_cos. Track divergence so the
Expand Down Expand Up @@ -578,14 +582,18 @@ void GgmlOvDecoder::add_extra_inputs() {
}
};

create_1d_input("attention_size", m_compute_params.attention_size);
if (m_compute_params.attention_size != -1) {
create_1d_input("attention_size", m_compute_params.attention_size);
}
if (m_compute_params.attention_size_swa != -1) {
create_1d_input("attention_size_swa", m_compute_params.attention_size_swa);
}
create_1d_input("n_seq_active", m_compute_params.n_seq_active);
create_1d_input("seq_active_start", m_compute_params.seq_active_start);
create_1d_input("seq_active_end", m_compute_params.seq_active_start + m_compute_params.n_seq_active);
create_1d_input("token_len_per_seq", m_compute_params.token_len_per_seq);
if (m_compute_params.token_len_per_seq != -1) {
create_1d_input("token_len_per_seq", m_compute_params.token_len_per_seq);
}
// create_1d_input("token_len", m_token_len_per_seq * m_n_seq_active);
}

Expand Down
Loading