Build Mega Service of AvatarChatbot on Gaudi

This document outlines the deployment process for a AvatarChatbot application utilizing the GenAIComps microservice pipeline on Intel Gaudi server.

🚀 Build Docker images

1. Source Code install GenAIComps

git clone https://github.com/opea-project/GenAIComps.git
cd GenAIComps

2. Build ASR Image

docker build -t opea/whisper-gaudi:latest --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy -f comps/asr/whisper/dependency/Dockerfile.intel_hpu .


docker build -t opea/asr:latest --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy -f comps/asr/whisper/Dockerfile .

3. Build LLM Image

docker build --no-cache -t opea/llm-tgi:latest --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy -f comps/llms/text-generation/tgi/Dockerfile .

4. Build TTS Image

docker build -t opea/speecht5-gaudi:latest --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy -f comps/tts/speecht5/dependency/Dockerfile.intel_hpu .

docker build -t opea/tts:latest --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy -f comps/tts/speecht5/Dockerfile .

5. Build Animation Image

docker build -t opea/wav2lip-gaudi:latest --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy -f comps/animation/wav2lip/dependency/Dockerfile.intel_hpu .

docker build -t opea/animation:latest --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy -f comps/animation/wav2lip/Dockerfile .

6. Build MegaService Docker Image

To construct the Mega Service, we utilize the GenAIComps microservice pipeline within the audioqna.py Python script. Build the MegaService Docker image using the command below:

git clone https://github.com/opea-project/GenAIExamples.git
cd GenAIExamples/AvatarChatbot/
docker build --no-cache -t opea/avatarchatbot:latest --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy -f Dockerfile .

Then run the command docker images, you will have following images ready:

  1. opea/whisper-gaudi:latest

  2. opea/asr:latest

  3. opea/llm-tgi:latest

  4. opea/speecht5-gaudi:latest

  5. opea/tts:latest

  6. opea/wav2lip-gaudi:latest

  7. opea/animation:latest

  8. opea/avatarchatbot:latest

🚀 Set the environment variables

Before starting the services with docker compose, you have to recheck the following environment variables.

export HUGGINGFACEHUB_API_TOKEN=<your_hf_token>
export host_ip=$(hostname -I | awk '{print $1}')

export TGI_LLM_ENDPOINT=http://$host_ip:3006
export LLM_MODEL_ID=Intel/neural-chat-7b-v3-3

export ASR_ENDPOINT=http://$host_ip:7066
export TTS_ENDPOINT=http://$host_ip:7055
export WAV2LIP_ENDPOINT=http://$host_ip:7860

export MEGA_SERVICE_HOST_IP=${host_ip}
export ASR_SERVICE_HOST_IP=${host_ip}
export TTS_SERVICE_HOST_IP=${host_ip}
export LLM_SERVICE_HOST_IP=${host_ip}
export ANIMATION_SERVICE_HOST_IP=${host_ip}

export MEGA_SERVICE_PORT=8888
export ASR_SERVICE_PORT=3001
export TTS_SERVICE_PORT=3002
export LLM_SERVICE_PORT=3007
export ANIMATION_SERVICE_PORT=3008

  • Gaudi2 HPU

export DEVICE="hpu"
export WAV2LIP_PORT=7860
export INFERENCE_MODE='wav2lip_only'
export CHECKPOINT_PATH='/usr/local/lib/python3.10/dist-packages/Wav2Lip/checkpoints/wav2lip_gan.pth'
export FACE="assets/img/avatar1.jpg"
# export AUDIO='assets/audio/eg3_ref.wav' # audio file path is optional, will use base64str in the post request as input if is 'None'
export AUDIO='None'
export FACESIZE=96
export OUTFILE="/outputs/result.mp4"
export GFPGAN_MODEL_VERSION=1.4 # latest version, can roll back to v1.3 if needed
export UPSCALE_FACTOR=1
export FPS=10

🚀 Start the MegaService

cd GenAIExamples/AvatarChatbot/docker_compose/intel/hpu/gaudi/
docker compose -f compose.yaml up -d

🚀 Test MicroServices

# whisper service
curl http://${host_ip}:7066/v1/asr \
  -X POST \
  -d '{"audio": "UklGRigAAABXQVZFZm10IBIAAAABAAEARKwAAIhYAQACABAAAABkYXRhAgAAAAEA"}' \
  -H 'Content-Type: application/json'

# asr microservice
curl http://${host_ip}:3001/v1/audio/transcriptions \
  -X POST \
  -d '{"byte_str": "UklGRigAAABXQVZFZm10IBIAAAABAAEARKwAAIhYAQACABAAAABkYXRhAgAAAAEA"}' \
  -H 'Content-Type: application/json'

# tgi service
curl http://${host_ip}:3006/generate \
  -X POST \
  -d '{"inputs":"What is Deep Learning?","parameters":{"max_new_tokens":17, "do_sample": true}}' \
  -H 'Content-Type: application/json'

# llm microservice
curl http://${host_ip}:3007/v1/chat/completions\
  -X POST \
  -d '{"query":"What is Deep Learning?","max_tokens":17,"top_k":10,"top_p":0.95,"typical_p":0.95,"temperature":0.01,"repetition_penalty":1.03,"streaming":false}' \
  -H 'Content-Type: application/json'

# speecht5 service
curl http://${host_ip}:7055/v1/tts \
  -X POST \
  -d '{"text": "Who are you?"}' \
  -H 'Content-Type: application/json'

# tts microservice
curl http://${host_ip}:3002/v1/audio/speech \
  -X POST \
  -d '{"text": "Who are you?"}' \
  -H 'Content-Type: application/json'

# wav2lip service
cd ../../../..
curl http://${host_ip}:7860/v1/wav2lip \
  -X POST \
  -d @assets/audio/sample_minecraft.json \
  -H 'Content-Type: application/json'

# animation microservice
curl http://${host_ip}:3008/v1/animation \
  -X POST \
  -d @assets/audio/sample_question.json \
  -H "Content-Type: application/json"

🚀 Test MegaService

curl http://${host_ip}:3009/v1/avatarchatbot \
  -X POST \
  -d @assets/audio/sample_whoareyou.json \
  -H 'Content-Type: application/json'

If the megaservice is running properly, you should see the following output:

"/outputs/result.mp4"

The output file will be saved in the current working directory, as ${PWD} is mapped to /outputs inside the wav2lip-service Docker container.

Gradio UI

sudo apt update
sudo apt install -y yasm pkg-config libx264-dev nasm
cd $WORKPATH
git clone https://github.com/FFmpeg/FFmpeg.git
cd FFmpeg
sudo ./configure --enable-gpl --enable-libx264 && sudo make -j$(nproc-1) && sudo make install && hash -r
pip install gradio==4.38.1 soundfile

cd $WORKPATH/GenAIExamples/AvatarChatbot
python3 ui/gradio/app_gradio_demo_avatarchatbot.py

The UI can be viewed at http://${host_ip}:7861
UI Example
In the current version v1.0, you need to set the avatar figure image/video and the DL model choice in the environment variables before starting AvatarChatbot backend service and running the UI. Please just customize the audio question in the UI.
** We will enable change of avatar figure between runs in v2.0

Troubleshooting

cd GenAIExamples/AvatarChatbot/tests
export IMAGE_REPO="opea"
export IMAGE_TAG="latest"
export HUGGINGFACEHUB_API_TOKEN=<your_hf_token>

test_avatarchatbot_on_gaudi.sh