Skip to content

Update jni runner #8978

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
merged 1 commit into from
Mar 7, 2025
Merged

Update jni runner #8978

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
28 changes: 21 additions & 7 deletions extension/android/jni/jni_layer_llama.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -132,16 +132,22 @@ class ExecuTorchLlamaJni
jint model_type_category,
facebook::jni::alias_ref<jstring> model_path,
facebook::jni::alias_ref<jstring> tokenizer_path,
jfloat temperature) {
jfloat temperature,
facebook::jni::alias_ref<jstring> data_path) {
return makeCxxInstance(
model_type_category, model_path, tokenizer_path, temperature);
model_type_category,
model_path,
tokenizer_path,
temperature,
data_path);
}

ExecuTorchLlamaJni(
jint model_type_category,
facebook::jni::alias_ref<jstring> model_path,
facebook::jni::alias_ref<jstring> tokenizer_path,
jfloat temperature) {
jfloat temperature,
facebook::jni::alias_ref<jstring> data_path = nullptr) {
#if defined(ET_USE_THREADPOOL)
// Reserve 1 thread for the main thread.
uint32_t num_performant_cores =
Expand All @@ -160,10 +166,18 @@ class ExecuTorchLlamaJni
tokenizer_path->toStdString().c_str(),
temperature);
} else if (model_type_category == MODEL_TYPE_CATEGORY_LLM) {
runner_ = std::make_unique<example::Runner>(
model_path->toStdString().c_str(),
tokenizer_path->toStdString().c_str(),
temperature);
if (data_path != nullptr) {
runner_ = std::make_unique<example::Runner>(
model_path->toStdString().c_str(),
tokenizer_path->toStdString().c_str(),
temperature,
data_path->toStdString().c_str());
} else {
runner_ = std::make_unique<example::Runner>(
model_path->toStdString().c_str(),
tokenizer_path->toStdString().c_str(),
temperature);
}
#if defined(EXECUTORCH_BUILD_MEDIATEK)
} else if (model_type_category == MODEL_TYPE_MEDIATEK_LLAMA) {
runner_ = std::make_unique<MTKLlamaRunner>(
Expand Down
16 changes: 12 additions & 4 deletions extension/android/src/main/java/org/pytorch/executorch/LlamaModule.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,16 +39,24 @@ public class LlamaModule {

@DoNotStrip
private static native HybridData initHybrid(
int modelType, String modulePath, String tokenizerPath, float temperature);
int modelType, String modulePath, String tokenizerPath, float temperature, String dataPath);

/** Constructs a LLAMA Module for a model with given path, tokenizer, and temperature. */
/** Constructs a LLAMA Module for a model with given model path, tokenizer, temperature. */
public LlamaModule(String modulePath, String tokenizerPath, float temperature) {
mHybridData = initHybrid(MODEL_TYPE_TEXT, modulePath, tokenizerPath, temperature);
mHybridData = initHybrid(MODEL_TYPE_TEXT, modulePath, tokenizerPath, temperature, null);
}

/**
* Constructs a LLAMA Module for a model with given model path, tokenizer, temperature and data
* path.
*/
public LlamaModule(String modulePath, String tokenizerPath, float temperature, String dataPath) {
mHybridData = initHybrid(MODEL_TYPE_TEXT, modulePath, tokenizerPath, temperature, dataPath);
}

/** Constructs a LLM Module for a model with given path, tokenizer, and temperature. */
public LlamaModule(int modelType, String modulePath, String tokenizerPath, float temperature) {
mHybridData = initHybrid(modelType, modulePath, tokenizerPath, temperature);
mHybridData = initHybrid(modelType, modulePath, tokenizerPath, temperature, null);
}

public void resetNative() {
Expand Down
Loading