v3.3.3

📅 Jun 18, 2025📦 tgiView on GitHub →

✨ 1 features🐛 4 fixes🔧 1 symbols

Summary

This release focuses on updating the Neuron backend, including bumping the SDK version and adding support for the Qwen3_moe model on Gaudi. Several Gaudi-specific fixes and performance optimizations were also implemented.

✨ New Features

[Gaudi] Enable Qwen3_moe model support.

🐛 Bug Fixes

[gaudi] Fixed an issue in benchmark tests related to Vlm rebase.
[Gaudi] Fixed an issue specific to the HuggingFaceM4/idefics2-8b model.
[Gaudi] Fixed integration test issues.
Used pad_token_id to pad input id for Gaudi.

🔧 Affected Symbols

get_cos_sin