Skip to content

Add a mock perf test for llama2 on Android #2963

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.

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

Closed
wants to merge 1 commit into from
Closed
Changes from all commits
Commits
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
Original file line number Diff line number Diff line change
@@ -0,0 +1,65 @@
/*
* Copyright (c) Meta Platforms, Inc. and affiliates.
* All rights reserved.
*
* This source code is licensed under the BSD-style license found in the
* LICENSE file in the root directory of this source tree.
*/

package com.example.executorchllamademo;

import static junit.framework.TestCase.assertTrue;
import static org.junit.Assert.assertEquals;
import static org.junit.Assert.assertFalse;

import androidx.test.ext.junit.runners.AndroidJUnit4;
import java.util.ArrayList;
import java.util.List;
import org.junit.Test;
import org.junit.runner.RunWith;
import org.pytorch.executorch.LlamaCallback;
import org.pytorch.executorch.LlamaModule;

@RunWith(AndroidJUnit4.class)
public class PerfTest implements LlamaCallback {

private static final String RESOURCE_PATH = "/data/local/tmp/llama/";
private static final String MODEL_NAME = "xnnpack_llama2.pte";
private static final String TOKENIZER_BIN = "tokenizer.bin";

// From https://github.com/pytorch/executorch/blob/main/examples/models/llama2/README.md
private static final Float EXPECTED_TPS = 7.0F;

private final List<String> results = new ArrayList<>();
private final List<Float> tokensPerSecond = new ArrayList<>();

@Test
public void testTokensPerSecond() {
String modelPath = RESOURCE_PATH + MODEL_NAME;
String tokenizerPath = RESOURCE_PATH + TOKENIZER_BIN;
LlamaModule mModule = new LlamaModule(modelPath, tokenizerPath, 0.8f);

int loadResult = mModule.load();
// Check that the model can be load successfully
assertEquals(0, loadResult);

// Run a testing prompt
mModule.generate("How do you do! I'm testing llama2 on mobile device", PerfTest.this);
assertFalse(tokensPerSecond.isEmpty());

final Float tps = tokensPerSecond.get(tokensPerSecond.size() - 1);
assertTrue(
"The observed TPS " + tps + " is less than the expected TPS " + EXPECTED_TPS,
tps >= EXPECTED_TPS);
}

@Override
public void onResult(String result) {
results.add(result);
}

@Override
public void onStats(float tps) {
tokensPerSecond.add(tps);
}
}