Create a new LLaMAPerfBenchmark app

pytorch · Aug 29, 2024 · 1cf1cc8 · 1cf1cc8
1 parent 4c17665
commit 1cf1cc8
Show file tree

Hide file tree

Showing 7 changed files with 419 additions and 62 deletions.
diff --git a/.github/workflows/apple-perf.yml b/.github/workflows/apple-perf.yml
@@ -265,9 +265,6 @@ jobs:
         curl "https://gha-artifacts.s3.amazonaws.com/${{ github.repository }}/${{ github.run_id }}/artifact/${{ matrix.model }}_${{ matrix.delegate }}/model.zip" -o model.zip
         unzip model.zip
 
-        # DEBUG: TO BE REMOVED
-        ls -lah
-
         ${CONDA_RUN} --no-capture-output \
           build/build_apple_llm_demo.sh ${{ matrix.tokenizer }} ${ARTIFACTS_DIR_NAME}
 

diff --git a/build/build_apple_llm_demo.sh b/build/build_apple_llm_demo.sh
@@ -20,7 +20,7 @@ cp *.pte *.bin "${TEST_RESOURCE_PATH}/"
 if [[ "${TOKENIZER}" = "bpe" ]]; then
   xcodebuild build-for-testing \
     -project "${APP_PATH}.xcodeproj" \
-    -scheme LLaMA \
+    -scheme LLaMAPerfBenchmark \
     -destination platform="iOS" \
     -allowProvisioningUpdates \
     DEVELOPMENT_TEAM=78E7V7QP35 \
@@ -33,7 +33,7 @@ if [[ "${TOKENIZER}" = "bpe" ]]; then
 else
   xcodebuild build-for-testing \
     -project "${APP_PATH}.xcodeproj" \
-    -scheme LLaMA \
+    -scheme LLaMAPerfBenchmark \
     -destination platform="iOS" \
     -allowProvisioningUpdates \
     DEVELOPMENT_TEAM=78E7V7QP35 \
@@ -53,7 +53,7 @@ PLATFORM="iphoneos"
 pushd "${BUILD_DIR}/${MODE}-${PLATFORM}"
 
 rm -rf Payload && mkdir Payload
-APP_NAME=LLaMA
+APP_NAME=LLaMAPerfBenchmark
 
 ls -lah
 cp -r "${APP_NAME}.app" Payload && zip -vr "${APP_NAME}.ipa" Payload

diff --git a/examples/demo-apps/apple_ios/LLaMA/LLaMA.xcodeproj/project.pbxproj b/examples/demo-apps/apple_ios/LLaMA/LLaMA.xcodeproj/project.pbxproj
diff --git a/examples/demo-apps/apple_ios/LLaMA/LLaMAPerfBenchmark/LLaMAPerfBenchmark.entitlements b/examples/demo-apps/apple_ios/LLaMA/LLaMAPerfBenchmark/LLaMAPerfBenchmark.entitlements
@@ -0,0 +1,8 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
+<plist version="1.0">
+<dict>
+	<key>com.apple.developer.kernel.increased-memory-limit</key>
+	<true/>
+</dict>
+</plist>
diff --git a/examples/demo-apps/apple_ios/LLaMA/LLaMAPerfBenchmark/LLaMAPerfBenchmarkApp.swift b/examples/demo-apps/apple_ios/LLaMA/LLaMAPerfBenchmark/LLaMAPerfBenchmarkApp.swift
@@ -0,0 +1,16 @@
+/*
+ * 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.
+ */
+
+import SwiftUI
+
+@main
+struct LLaMAPerfBenchmarkApp: SwiftUI.App {
+  var body: some Scene {
+    WindowGroup { }
+  }
+}
diff --git a/...ple_ios/LLaMA/LLaMATests/LLaMATests.swift → ...chmarkTests/LLaMAPerfBenchmarkTests.swift b/...ple_ios/LLaMA/LLaMATests/LLaMATests.swift → ...chmarkTests/LLaMAPerfBenchmarkTests.swift
@@ -9,7 +9,7 @@
 import XCTest
 import LLaMARunner
 
-final class LLaMATests: XCTestCase {
+final class LLaMAPerfBenchmarkTests: XCTestCase {
   func testLlama2() throws {
     guard
       let modelPath = Bundle(for: type(of: self))
@@ -34,5 +34,12 @@ final class LLaMATests: XCTestCase {
       XCTFail("Failed to load the model: \(loadError)")
     }
     XCTAssertTrue(runner.isloaded())
+
+    let seq_len = 128
+    var tokens: [String] = []
+    try runner.generate("How do you do! I'm testing llama2 on mobile device", sequenceLength: seq_len) { token in
+      tokens.append(token)
+    }
   }
 }
+
diff --git a/...ios/LLaMA/LLaMATests/Resources/.gitignore → ...MAPerfBenchmarkTests/Resources/.gitignore b/...ios/LLaMA/LLaMATests/Resources/.gitignore → ...MAPerfBenchmarkTests/Resources/.gitignore