AI BriefingGaudi 2 Accelerates Protein Model ProtST
Key point
Intel Gaudi 2 accelerators significantly improved the inference and fine-tuning performance of the protein language model ProtST compared to A100.
Details
MILA and Intel Labs have released an optimization method for efficiently running and fine-tuning the multimodal protein design model ProtST on Intel Gaudi 2 accelerators.
Using the Optimum for Intel Gaudi open-source library, existing Transformers-based scripts can be ported to Gaudi 2 with only minimal code changes.
Performance comparison results versus NVIDIA A100 80GB:
- Inference: Gaudi 2 recorded 1.76x faster speed than A100.
- Fine-tuning: Gaudi 2 showed 2.92x faster speed than A100, and demonstrated near-linear scalability in distributed training using 4 or 8 accelerators.
The ProtST model and related datasets are available through the Hugging Face Hub.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Qwen3.5-based Small Decision Model 'Kev' Released with TypeSafe API Compatibility and Performance Metrics