llama.cpp adds tensor split fix for Intel GPUs
The b9788 release includes a fix for . The change adds SYCL support for the `--split-mode tensor` option. This matters when a model’s work is split across more than one GPU at the tensor level.
No speed, memory, or cost numbers are available yet. People with two are being asked to test it and share results.
Key points
- release b9788 includes an Intel GPU-related fix.
- SYCL now supports the `--split-mode tensor` option.
- The change is relevant to splitting model work across multiple GPUs.
- No numbers have been shared yet.
- Testing on two would show whether it helps in practice.