ggml/llama.cpp releases: b10455

TL;DR

The ggml/llama.cpp project released version b10455, which includes updates and fixes for SYCL support, particularly for optimization steps in AdamW and SGD algorithms. These changes are important for enhancing the project's performance and compatibility across different computing environments.

Detailed Summary

The ggml/llama.cpp project released a new version (b10455) that includes support for OP OPT_STEP_ADAMW and OPT_STEP_SGD, along with fixes to conflicts in ops.md. This update was co-authored by Neo Zhang Jianyu from Intel. The broader impact is reflected in the project's ongoing development of optimization steps for machine learning models, potentially enhancing performance across various platforms including macOS/iOS.

Key Points

• • [SYCL] support OP OPT_STEP_ADAMW, OPT_STEP_SGD
• • Fix conflict of ops.md
• • Update the ops.md