dpas

Here are 2 public repositories matching this topic...

Attention and multi-GPU work for the Intel Arc A770 (Xe-HPG) in SYCL, for llama.cpp. Sibling of volta-fa.

gpgpu sycl gpu-kernels oneapi xmx a770 intel-arc llama-cpp llm-inference dpas xe-hpg

Native Intel Xe2 LLM inference: SYCL/DPAS kernels, paged KV, prefix caching, continuous batching, and XMX packed prefill

intel sycl xe2 oneapi level-zero xmx intel-xpu intel-arc llm-inference qwen paged-attention dpas continuous-batching openai-compatible prefix-caching

To associate your repository with the dpas topic, visit your repo's landing page and select "manage topics."