-
Notifications
You must be signed in to change notification settings - Fork 29
167 lines (152 loc) · 6.34 KB
/
Copy pathllm-android.yml
File metadata and controls
167 lines (152 loc) · 6.34 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
# 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.
name: LlamaDemo Android
on:
pull_request:
branches: [main]
paths:
- 'llm/android/**'
- '.github/workflows/llm-android.yml'
workflow_dispatch:
inputs:
model_preset:
description: 'Model preset to use'
required: true
type: choice
options:
- stories
- llama
- qwen3
- custom
default: 'stories'
custom_pte_url:
description: 'Custom URL for model .pte file (only used when model_preset is custom)'
required: false
type: string
custom_tokenizer_url:
description: 'Custom URL for tokenizer file (only used when model_preset is custom)'
required: false
type: string
permissions:
contents: read
jobs:
instrumentation-test:
runs-on: 8-core-ubuntu
env:
API_LEVEL: 34
ARCH: x86_64
EMULATOR_OPTIONS: -no-window -gpu swiftshader_indirect -noaudio -no-boot-anim -camera-back none
RAM_SIZE: 16384
name: Instrumentation Test LlamaDemo
steps:
- name: Checkout repository
uses: actions/checkout@v4
- name: Write job summary
run: |
echo "## Test Configuration" >> $GITHUB_STEP_SUMMARY
echo "| Parameter | Value |" >> $GITHUB_STEP_SUMMARY
echo "|-----------|-------|" >> $GITHUB_STEP_SUMMARY
echo "| Model Preset | \`${{ inputs.model_preset || 'stories' }}\` |" >> $GITHUB_STEP_SUMMARY
if [ "${{ inputs.model_preset }}" = "custom" ]; then
echo "| Custom PTE URL | \`${{ inputs.custom_pte_url }}\` |" >> $GITHUB_STEP_SUMMARY
echo "| Custom Tokenizer URL | \`${{ inputs.custom_tokenizer_url }}\` |" >> $GITHUB_STEP_SUMMARY
fi
- name: Enable KVM group perms
run: |
echo 'KERNEL=="kvm", GROUP="kvm", MODE="0666", OPTIONS+="static_node=kvm"' | sudo tee /etc/udev/rules.d/99-kvm4all.rules
sudo udevadm control --reload-rules
sudo udevadm trigger --name-match=kvm
- name: Set up JDK 17
uses: actions/setup-java@v4
with:
java-version: '17'
distribution: 'temurin'
- name: Setup Gradle
uses: gradle/actions/setup-gradle@v4
- name: AVD cache
uses: actions/cache@v4
id: avd-cache
with:
path: |
~/.android/avd/*
~/.android/adb*
key: avd-${{ env.API_LEVEL }}-${{ env.ARCH }}-ram${{ env.RAM_SIZE }}
- name: Create AVD and generate snapshot for caching
if: steps.avd-cache.outputs.cache-hit != 'true'
uses: reactivecircus/android-emulator-runner@v2
with:
api-level: ${{ env.API_LEVEL }}
arch: ${{ env.ARCH }}
force-avd-creation: true
emulator-options: ${{ env.EMULATOR_OPTIONS }}
disable-animations: false
working-directory: llm/android/LlamaDemo
script: echo "Generated AVD snapshot for caching."
- name: Configure AVD RAM
run: |
AVD_DIR="$HOME/.android/avd"
for config in "$AVD_DIR"/*.avd/config.ini; do
if [ -f "$config" ]; then
echo "Updating RAM in $config"
sed -i 's/hw.ramSize=.*/hw.ramSize=${{ env.RAM_SIZE }}/' "$config" || true
grep -q "hw.ramSize" "$config" || echo "hw.ramSize=${{ env.RAM_SIZE }}" >> "$config"
fi
done
- name: Run instrumentation tests
uses: reactivecircus/android-emulator-runner@v2
env:
MODEL_PRESET: ${{ inputs.model_preset || 'stories' }}
CUSTOM_PTE_URL: ${{ inputs.custom_pte_url }}
CUSTOM_TOKENIZER_URL: ${{ inputs.custom_tokenizer_url }}
with:
api-level: ${{ env.API_LEVEL }}
arch: ${{ env.ARCH }}
force-avd-creation: false
emulator-options: -no-snapshot-save ${{ env.EMULATOR_OPTIONS }}
disable-animations: true
working-directory: llm/android/LlamaDemo
script: |
adb shell rm -rf /data/local/tmp/llama
adb shell mkdir -p /data/local/tmp/llama
echo "=== Model directory after cleanup ==="
adb shell ls -la /data/local/tmp/llama/ || echo "Directory empty or not found"
adb logcat -c && adb logcat > /tmp/logcat.txt &
LOGCAT_PID=$!
if [ "$MODEL_PRESET" = "custom" ]; then GRADLE_ARGS="-PmodelPreset=$MODEL_PRESET -PcustomPteUrl=$CUSTOM_PTE_URL -PcustomTokenizerUrl=$CUSTOM_TOKENIZER_URL"; else GRADLE_ARGS="-PmodelPreset=$MODEL_PRESET"; fi
echo "=== Running Gradle with preset: $MODEL_PRESET ==="
./gradlew connectedCheck $GRADLE_ARGS; TEST_EXIT_CODE=$?
echo "=== Model directory after Gradle ==="
adb shell ls -la /data/local/tmp/llama/
kill $LOGCAT_PID || true
# Show which model was used
echo "=== Model configuration used by test ==="
grep "UIWorkflowTest.*Using model" /tmp/logcat.txt || echo "Model config not found in logcat"
# Extract response from logcat
echo "=== Searching for LLAMA_RESPONSE in logcat ==="
grep "LLAMA_RESPONSE" /tmp/logcat.txt || echo "No LLAMA_RESPONSE found in logcat"
grep "LLAMA_RESPONSE" /tmp/logcat.txt | sed 's/.*LLAMA_RESPONSE: //' | grep -v "BEGIN_RESPONSE\|END_RESPONSE" > /tmp/response.txt || true
echo "=== Response file contents ==="
cat /tmp/response.txt || echo "Response file empty or not created"
# Clean up model files to avoid caching large files
adb shell rm -rf /data/local/tmp/llama
exit $TEST_EXIT_CODE
- name: Add model response to summary
if: always()
run: |
if [ -f /tmp/response.txt ]; then
echo "" >> $GITHUB_STEP_SUMMARY
echo "## Model Response" >> $GITHUB_STEP_SUMMARY
echo '```' >> $GITHUB_STEP_SUMMARY
cat /tmp/response.txt >> $GITHUB_STEP_SUMMARY
echo '```' >> $GITHUB_STEP_SUMMARY
fi
- name: Upload logcat
if: always()
uses: actions/upload-artifact@v4
with:
name: logcat
path: /tmp/logcat.txt
retention-days: 7