Skip to content

Add kv data type flag #949

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
Jul 9, 2026
Merged

Add kv data type flag #949

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
16 changes: 16 additions & 0 deletions gemma/gemma_args.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -211,6 +211,7 @@ struct InferenceArgs : public ArgsBase<InferenceArgs> {
Path prompt_file;
std::string eot_line;
std::string attention_impl;
std::string kv_cache_type;

template <class Visitor>
void ForEach(const Visitor& visitor) {
Expand Down Expand Up @@ -266,6 +267,10 @@ struct InferenceArgs : public ArgsBase<InferenceArgs> {
2);
visitor(attention_impl, "attention_impl", std::string("flash"),
"Attention implementation to use. See configs.cc for options.", 2);
visitor(kv_cache_type, "kv_cache_type", std::string(""),
"KV cache data type (f32, bf16, int8). If empty, deduced from "
"attention_impl.",
2);
}

void CopyTo(RuntimeConfig& runtime_config) const {
Expand All @@ -288,6 +293,17 @@ struct InferenceArgs : public ArgsBase<InferenceArgs> {
runtime_config.temperature = temperature;
runtime_config.top_k = top_k;
runtime_config.attention_impl = GetAttentionImpl(attention_impl);
if (!kv_cache_type.empty()) {
if (kv_cache_type == "int8" || kv_cache_type == "i8") {
runtime_config.kv_cache_type = Type::kInt8;
} else if (kv_cache_type == "bf16") {
runtime_config.kv_cache_type = Type::kBF16;
} else if (kv_cache_type == "f32" || kv_cache_type == "float") {
runtime_config.kv_cache_type = Type::kF32;
} else {
HWY_ABORT("Unknown kv_cache_type: %s\n", kv_cache_type.c_str());
}
}
}
};

Expand Down
Loading