Simple LLM Inference: Interactive Text Generation with Intel GPUs

LLM Inference example is a Python script designed to demonstrate interactive text generation using pre-trained Language Models (LLMs) from Hugging Face Transformers and Intel dGPUs. Although the code structure resembles a chatbot, it's important to note that the models used were not specifically trained for conversational purposes. This repository provides a Command Line Interface (CLI) to interact with the models in two modes: with context (remembering previous interactions) and without context. The code is optimized to run on Intel GPUs using Intel Extension for PyTorch (IPEX).

Features

• Model Selection: Choose between predefined models or enter a custom model repository from Hugging Face Hub.
• Context Control: Interact with the model in two modes, with and without context.
• Generation Parameters Control: Customize the response generation by adjusting parameters like temperature, top_p, top_k, num_beams, and repetition_penalty.
• Repetition Removal: The code includes logic to remove repetitive sentences from the generated text.

Prerequisites

• Python 3.6 or higher
• Intel Extension for PyTorch (IPEX)
• Hugging Face Transformers
• Hugging Face Accelerate

Setup

1. Clone the repository

git clone https://github.com/rahulunair/simple_llm_inference.git && cd simple_llm_inference

2. Install dependencies:

python -m pip install torch==2.0.1a0 torchvision==0.15.2a0 intel_extension_for_pytorch==2.0.110+xpu -f https://developer.intel.com/ipex-whl-stable-xpu
python -m pip install transformers accelerate sentencepiece

To know more about IPEX on Intel GPUs use the following guide.

Usage

1. Run bot.py

python bot.py

2. Follow the on-screen prompts to select a model and the interaction mode.

Sample output

Please select a model:
1. Writer/camel-5b-hf
2. openlm-research/open_llama_3b_v2
3. Enter a custom model repo from HuggingFace Hub
Enter 1 to 3: 2
Using max length: 256
Note: This is a demonstration using pretrained models which were not fine-tuned for chat.
You can choose between two modes of interaction:
1. Interact with context
2. Interact without context
Enter 1 or 2: 1
You: Hello, Bot!
Bot: Hello! How can I assist you today?

Contributing

Feel free to submit issues and pull requests. Contributions are welcome!