Skip to content

dflash: let the log-SNR input opt into graph reuse #124

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
khosravipasha merged 1 commit into from
Aug 27, 2026
Merged

dflash: let the log-SNR input opt into graph reuse #124

Show file tree
Hide file tree
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
6 changes: 6 additions & 0 deletions src/llama-graph.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -75,6 +75,12 @@ void llm_graph_input_dspark_logsnr::set_input(const llama_ubatch * ubatch) {
}
}

bool llm_graph_input_dspark_logsnr::can_reuse(const llm_graph_params & params) {
// v_feat is a function of n_tokens and n_seqs_unq, both already checked by
// llm_graph_params::allow_reuse, and of min/max_log_snr, fixed per model
return feat && feat->ne[1] == params.ubatch.n_tokens;
}

void llm_graph_input_embd::set_input(const llama_ubatch * ubatch) {
if (ubatch->token) {
const int64_t n_tokens = ubatch->n_tokens;
Expand Down
2 changes: 2 additions & 0 deletions src/llama-graph.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -152,6 +152,8 @@ class llm_graph_input_dspark_logsnr : public llm_graph_input_i {

void set_input(const llama_ubatch * ubatch) override;

bool can_reuse(const llm_graph_params & params) override;

ggml_tensor * feat = nullptr; // F32 [128, n_tokens]

std::vector<float> v_feat;
Expand Down