Voice Assistant

In this directory, you will find examples on how you could apply BigDL-LLM INT4 optimizations on Whisper and Llama2 models on Intel GPUs. For illustration purposes, we utilize the following models:

openai/whisper-small and openai/whisper-medium as reference whisper models.
meta-llama/Llama-2-7b-chat-hf and meta-llama/Llama-2-13b-chat-hf as reference Llama2 models.

0. Requirements

To run these examples with BigDL-LLM on Intel GPUs, we have some recommended requirements for your machine, please refer to here for more information.

Example: Predict Tokens using `generate()` API

In the example generate.py, we show a basic use case for a Whisper model to conduct transcription using generate() API, then use the recoginzed text as the input for Llama2 model to predict the next N tokens using generate() API, with BigDL-LLM INT4 optimizations on Intel GPUs.

1. Install

We suggest using conda to manage environment:

conda create -n llm python=3.9
conda activate llm
# below command will install intel_extension_for_pytorch==2.0.110+xpu as default
# you can install specific ipex/torch version for your need
pip install --pre --upgrade bigdl-llm[xpu] -f https://developer.intel.com/ipex-whl-stable-xpu
pip install librosa soundfile datasets
pip install accelerate
pip install SpeechRecognition sentencepiece colorama
# If you failed to install PyAudio, try to run sudo apt install portaudio19-dev on ubuntu
pip install PyAudio inquirer sounddevice

2. Configures OneAPI environment variables

source /opt/intel/oneapi/setvars.sh

3. Run

For optimal performance on Arc, it is recommended to set several environment variables.

export USE_XETLA=OFF
export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1

python ./generate.py --llama2-repo-id-or-model-path REPO_ID_OR_MODEL_PATH --whisper-repo-id-or-model-path REPO_ID_OR_MODEL_PATH --n-predict N_PREDICT

Arguments info:

--llama2-repo-id-or-model-path REPO_ID_OR_MODEL_PATH: argument defining the huggingface repo id for the Llama2 model (e.g. meta-llama/Llama-2-7b-chat-hf and meta-llama/Llama-2-13b-chat-hf) to be downloaded, or the path to the huggingface checkpoint folder. It is default to be 'meta-llama/Llama-2-7b-chat-hf'.
--whisper-repo-id-or-model-path REPO_ID_OR_MODEL_PATH: argument defining the huggingface repo id for the Whisper model (e.g. openai/whisper-small and openai/whisper-medium) to be downloaded, or the path to the huggingface checkpoint folder. It is default to be 'openai/whisper-small'.
--n-predict N_PREDICT: argument defining the max number of tokens to predict. It is default to be 32.

Known Issues

The speech_recognition library may occasionally skip recording due to low volume. An alternative option is to save the recording in WAV format using PyAudio and read the file as an input. Here is an example using PyAudio:

import pyaudio
import speech_recognition as sr

CHUNK = 1024
FORMAT = pyaudio.paInt16
CHANNELS = 1                # The desired number of input channels
RATE = 16000                # The desired rate (in Hz)
RECORD_SECONDS = 10         # Recording time (in second)
WAVE_OUTPUT_FILENAME = "/path/to/pyaudio_out.wav"
p = pyaudio.PyAudio()
                
stream = p.open(format=FORMAT,
                channels=CHANNELS,
                rate=RATE,
                input=True,
                frames_per_buffer=CHUNK)

print("*"*10, "Listening\n")
frames = []
data =0
for i in range(0, int(RATE / CHUNK * RECORD_SECONDS)):
  data = stream.read(CHUNK)  ## <class 'bytes'> ,exception_on_overflow = False
  frames.append(data)   ## <class 'list'>
print("*"*10, "Stop recording\n")

stream.stop_stream()
stream.close()
p.terminate()

wf = wave.open(WAVE_OUTPUT_FILENAME, 'wb')
wf.setnchannels(CHANNELS)
wf.setsampwidth(p.get_sample_size(FORMAT))
wf.setframerate(RATE)
wf.writeframes(b''.join(frames))
wf.close()

r = sr.Recognizer()
with sr.AudioFile(WAVE_OUTPUT_FILENAME) as source1:
    audio = r.record(source1)  # read the entire audio file   
frame_data = np.frombuffer(audio.frame_data, np.int16).flatten().astype(np.float32) / 32768.0

Sample Output

(llm) bigdl@bigdl-llm:~/Documents/voiceassistant$ python generate.py --llama2-repo-id-or-model-path /mnt/windows/demo/models/Llama-2-7b-chat-hf --whisper-repo-id-or-model-path /mnt/windows/demo/models/whisper-medium
/home/bigdl/anaconda3/envs/llm/lib/python3.9/site-packages/torchvision/io/image.py:13: UserWarning: Failed to load image Python extension: ''If you don't plan on using image functionality from `torchvision.io`, you can ignore this warning. Otherwise, there might be something wrong with your environment. Did you have `libjpeg` or `libpng` installed before building `torchvision` from source?
  warn(

[?] Which microphone do you choose?: Default
 > Default
   HDA Intel PCH: ALC274 Analog (hw:0,0)
   HDA Intel PCH: HDMI 0 (hw:0,3)
   HDA Intel PCH: HDMI 1 (hw:0,7)
   HDA Intel PCH: HDMI 2 (hw:0,8)
   HDA Intel PCH: HDMI 3 (hw:0,9)
   HDA Intel PCH: HDMI 4 (hw:0,10)
   HDA Intel PCH: HDMI 5 (hw:0,11)
   HDA Intel PCH: HDMI 6 (hw:0,12)
   HDA Intel PCH: HDMI 7 (hw:0,13)
   HDA Intel PCH: HDMI 8 (hw:0,14)
   HDA Intel PCH: HDMI 9 (hw:0,15)
   HDA Intel PCH: HDMI 10 (hw:0,16)

The device name Default is selected.
Downloading builder script: 100%|██████████████████████████████████████████████████████| 5.17k/5.17k [00:00<00:00, 14.3MB/s]
Downloading data: 100%|████████████████████████████████████████████████████████████████████████████████████████| 9.08M/9.08M [00:01<00:00, 4.75MB/s]
Downloading data files: 100%|████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:04<00:00,  4.57s/it]]
Extracting data files: 100%|██████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 39.98it/s]
Generating validation split: 73 examples [00:00, 5328.37 examples/s]
Converting and loading models...
Loading checkpoint shards: 100%|██████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:09<00:00,  3.04s/it]
/home/bigdl/anaconda3/envs/yina-llm/lib/python3.9/site-packages/transformers/generation/configuration_utils.py:362: UserWarning: `do_sample` is set to `False`. However, `temperature` is set to `0.9` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `temperature`. This was detected when initializing the generation config instance, which means the corresponding file may hold incorrect parameterization and should be fixed.
  warnings.warn(
/home/bigdl/anaconda3/envs/yina-llm/lib/python3.9/site-packages/transformers/generation/configuration_utils.py:367: UserWarning: `do_sample` is set to `False`. However, `top_p` is set to `0.6` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `top_p`. This was detected when initializing the generation config instance, which means the corresponding file may hold incorrect parameterization and should be fixed.
  warnings.warn(
/home/bigdl/anaconda3/envs/yina-llm/lib/python3.9/site-packages/transformers/generation/utils.py:1411: UserWarning: You have modified the pretrained model configuration to control generation. This is a deprecated strategy to control generation and will be removed soon, in a future version. Please use a generation configuration file (see https://huggingface.co/docs/transformers/main_classes/text_generation )
  warnings.warn(
Calibrating...
Listening now...
Recognizing...

Whisper : 
 What is AI?

BigDL-LLM: 
 Artificial intelligence (AI) is the broader field of research and development aimed at creating machines that can perform tasks that typically require human intelligence,
Listening now...
Recognizing...

Whisper : 
 Tell me something about Intel

BigDL-LLM: 
 Intel is a well-known technology company that specializes in designing, manufacturing, and selling computer hardware components and semiconductor products.

8.9 KiB Raw Blame History