Skip to content

Fix VLM prefill batch size - prompt+tokens #859

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
copybara-service merged 1 commit into from
Mar 5, 2026
Merged

Fix VLM prefill batch size - prompt+tokens #859

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
7 changes: 3 additions & 4 deletions gemma/run.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -200,15 +200,14 @@ void ReplGemma(const GemmaArgs& args, const Gemma& gemma, KVCache& kv_cache,
config.wrapping, abs_pos, prompt_string,
image_tokens.Rows());
runtime_config.image_tokens = &image_tokens;
// PrefixLM sees/attends to all tokens.
runtime_config.prefill_tbatch_size = prompt.size();

prompt_size = prompt.size() - image_tokens.Rows();
if (config.wrapping == PromptWrapping::PALIGEMMA) {
// The end of the prefix for prefix-LM style attention in Paligemma.
// See Figure 2 of https://arxiv.org/abs/2407.07726.
prefix_end = prompt_size;
// We need to look at all the tokens for the prefix.
// NOTE: Online softmax is on the roadmap, after which this requirement
// can be lifted.
runtime_config.prefill_tbatch_size = prompt_size;
}
} else {
prompt = WrapAndTokenize(gemma.Tokenizer(), gemma.ChatTemplate(),
Expand Down
Loading