Current ROCm MTP status

Use for the validated F16/off configuration:

  --pipeline-parallel off
  --cache-type-k-draft f16
  --cache-type-v-draft f16

The current complete patch is:

  bug/sol4.patch

It includes the MTP fitting and pipeline changes plus the merged HIP VEC dispatch from vec_sol. rocm_improvement.patch is an identical duplicate of sol4.patch.

The VEC Q4/Q4 MTP run initially reached 147,712 context tokens and completed startup plus short generation. Long-prompt stability, delayed-OOM, throughput, and draft-acceptance testing remain pending.

Validation notes:

- CPU/default build completed in an isolated master worktree.
- HIP compiled fattn.cu, all fattn-vec instances, and linked libggml-hip.so.
- The final HIP llama-server link was stopped before completion.
- The real master checkout was not modified.

For Vulkan, initially retain:

  --pipeline-parallel auto

Compare Q4_0 and F16 draft KV per model and monitor fitted context, pipeline mode, buffer sizes, allocation failures, throughput, draft acceptance, and long-prompt stability.

Session attribution: Sol

Historical notes and logs are under bug/logs/; archived patches are under bug/old/patches/.
