We're excited to announce the native integration of Intel Gaudi hardware support directly into Text Generation Inference (TGI), our production-ready serving solution for Large Language Models (LLMs). This integration brings the power of Intel's specialized AI accelerators to our high-performance inference stack, enabling more deployment options for the open-source AI community 🎉
We've fully integrated Gaudi support into TGI's main codebase in PR #3091. Previously, we maintained a separate fork for Gaudi devices at tgi-gaudi. This was cumbersome for users and prevented us from supporting the latest TGI features at launch. Now using the new TGI multi-backend architecture, we support Gaudi directly on TGI – no more finicking on a custom repository 🙌
This integration supports Intel's full line of Gaudi hardware:
Gaudi1 💻: Available on AWS EC2 DL1 instances
Gaudi2 💻💻: Available on Intel Tiber AI Cloud and Denvr Dataworks
Gaudi3 💻💻💻: Available on Intel Tiber AI Cloud, IBM Cloud and from OEM such as Dell, HP and Supermicro
You can also find more information on Gaudi hardware on Intel's Gaudi product page
The Gaudi backend for TGI provides several key benefits:
Hardware Diversity 🔄: More options for deploying LLMs in production beyond traditional GPUs
Cost Efficiency 💰: Gaudi hardware often provides compelling price-performance for specific workloads
Production-Ready ⚙️: All the robustness of TGI (dynamic batching, streamed responses, etc.) now available on Gaudi
Model Support 🤖: Run popular models like Llama 3.1, Mixtral, Mistral, and more on Gaudi hardware
Advanced Features 🔥: Support for multi-card inference (sharding), vision-language models, and FP8 precision
The easiest way to run TGI on Gaudi is to use our official Docker image. You need to run the image on a Gaudi hardware machine. Here's a basic example to get you started:
model=meta-llama/Meta-Llama-3.1-8B-Instruct
volume=$PWD/data # share a volume with the Docker container to avoid downloading weights every run
hf_token=YOUR_HF_ACCESS_TOKEN
docker run --runtime=habana --cap-add=sys_nice --ipc=host \
-p 8080:80 \
-v $volume:/data \
-e HF_TOKEN=$hf_token \
-e HABANA_VISIBLE_DEVICES=all \
ghcr.io/huggingface/text-generation-inference:3.2.1-gaudi \
--model-id $model Once the server is running, you can send inference requests:
curl 127.0.0.1:8080/generate
-X POST
-d '{"inputs":"What is Deep Learning?","parameters":{"max_new_tokens":32}}'
-H 'Content-Type: application/json'For comprehensive documentation on using TGI with Gaudi, including how-to guides and advanced configurations, refer to the new dedicated Gaudi backend documentation.
We have optimized the following models for both single and multi-card configurations. This means these models run as fast as possible on Intel Gaudi. We've specifically optimized the modeling code to target Intel Gaudi hardware, ensuring we offer the best performance and fully utilize Gaudi's capabilities:
Llama 3.1 (8B and 70B)
Llama 3.3 (70B)
Llama 3.2 Vision (11B)
Mistral (7B)
Mixtral (8x7B)
CodeLlama (13B)
Falcon (180B)
Qwen2 (72B)
Starcoder and Starcoder2
Gemma (7B)
Llava-v1.6-Mistral-7B
Phi-2
🏃♂️ We also offer many advanced features on Gaudi hardware, such as FP8 quantization thanks to Intel Neural Compressor (INC), enabling even greater performance optimizations.
✨ Coming soon! We're excited to expand our model lineup with cutting-edge additions including DeepSeek-r1/v3, QWen-VL, and more powerful models to power your AI applications! 🚀