March 2026: Delivered Intel XPU support in Docker deployment for ai-dynamo/dynamo, adding a Dockerfile and build workflow changes to enable deployment on Intel hardware alongside CUDA. This expands hardware compatibility and enables customers with Intel infrastructure to run deep learning tasks using the same codebase and CI/CD flow.

June 2025 maintenance-focused update for intel-analytics/ipex-llm. Implemented a critical robustness fix for BERT attention mask handling to ensure correct behavior when the mask is missing or entirely false, preventing potential errors in attention computation and improving inference reliability across edge cases. No new features were delivered this month; the primary emphasis was stability and correctness of the attention mechanism to reduce production incidents and support reliable model inference.