Skip to content

Latest commit

 

History

History
112 lines (79 loc) · 3.7 KB

README.md

File metadata and controls

112 lines (79 loc) · 3.7 KB

Build Mega Service of AudioQnA on Gaudi

This document outlines the deployment process for a AudioQnA 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/src/integrations/dependency/whisper/Dockerfile.intel_hpu .

3. Build LLM Image

Intel Xeon optimized image hosted in huggingface repo will be used for TGI service: ghcr.io/huggingface/tgi-gaudi:2.0.6 (https://github.com/huggingface/tgi-gaudi)

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/src/integrations/dependency/speecht5/Dockerfile.intel_hpu .

5. 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/AudioQnA/
docker build --no-cache -t opea/audioqna: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/speecht5-gaudi:latest
  3. opea/audioqna:latest

🚀 Set the environment variables

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

export host_ip=<your External Public IP>    # export host_ip=$(hostname -I | awk '{print $1}')
export HUGGINGFACEHUB_API_TOKEN=<your HF token>

export LLM_MODEL_ID=Intel/neural-chat-7b-v3-3

export MEGA_SERVICE_HOST_IP=${host_ip}
export WHISPER_SERVER_HOST_IP=${host_ip}
export SPEECHT5_SERVER_HOST_IP=${host_ip}
export LLM_SERVER_HOST_IP=${host_ip}

export WHISPER_SERVER_PORT=7066
export SPEECHT5_SERVER_PORT=7055
export LLM_SERVER_PORT=3006

export BACKEND_SERVICE_ENDPOINT=http://${host_ip}:3008/v1/audioqna

🚀 Start the MegaService

NOTE: Users will need at least three Gaudi cards for AudioQnA.

cd GenAIExamples/AudioQnA/docker_compose/intel/hpu/gaudi/
docker compose up -d

🚀 Test MicroServices

# whisper service
curl http://${host_ip}:7066/v1/asr \
  -X POST \
  -d '{"audio": "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'

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

🚀 Test MegaService

Test the AudioQnA megaservice by recording a .wav file, encoding the file into the base64 format, and then sending the base64 string to the megaservice endpoint. The megaservice will return a spoken response as a base64 string. To listen to the response, decode the base64 string and save it as a .wav file.

# voice can be "default" or "male"
curl http://${host_ip}:3008/v1/audioqna \
  -X POST \
  -d '{"audio": "UklGRigAAABXQVZFZm10IBIAAAABAAEARKwAAIhYAQACABAAAABkYXRhAgAAAAEA", "max_tokens":64, "voice":"default"}' \
  -H 'Content-Type: application/json' | sed 's/^"//;s/"$//' | base64 -d > output.wav