docs: Ling-3.0-flash Q6_K trial page (Vulkan+GTT memory research, switch-back steps)

2026-08-14 01:22:02 -07:00
parent 923e775354
commit c275cbaf57
1 changed files with 90 additions and 0 deletions
@@ -0,0 +1,90 @@
+# Ling-3.0-flash Q6_K Trial on Dankclaw
+
+This page records the Ling-3.0-flash trial that temporarily replaced
+`frank-default-large` on Dankclaw. It replaces the Qwen3.6-27B service on
+port 7770 under the same API alias.
+
+## Status
+
+- Service: `frank-large-ling` (systemd).
+- Model: `inclusionAI/Ling-3.0-flash` Q6_K (MoE, ~118B total params).
+- Runtime: llama.cpp **fork** `aetherbird/llama.cpp` branch
+  `bailingmoe3-support` (commit `db4480bc8`, build b10411) at
+  `/home/dank/llama.cpp-ling`. Stock llama.cpp cannot load the
+  `bailingmoe3` architecture.
+- Backend: `Vulkan0` (RADV) with AMD GTT.
+- Spec: `draft-mtp` `n_max 3`. The MTP block is bundled in the Ling GGUF
+  itself (block 42) - no separate drafter file.
+- Context: 262144 total, 2 slots x 131072, batch 512 / ubatch 256.
+- Measured: load ~45s, generation ~7.7 t/s (cold, 4k prompt) to ~36 t/s
+  (tiny cached prompt). Concurrent streams both complete.
+
+## Why Vulkan + GTT
+
+The Strix Halo APU has 8 x 16 GB LPDDR5 modules = **128 GB physical
+shared memory**. The firmware carves ~96 GiB as dedicated iGPU VRAM; the
+kernel sees the remainder (~30 GiB, `MemTotal`). GRUB adds no memory
+options (`iommu=pt` only).
+
+The ROCm runtime reports 96 GiB VRAM. **Vulkan (RADV) reports a
+111.2 GiB heap** (113910 MiB) because it can also reach system memory
+through the GTT. For a model near 100 GiB, Vulkan is the only backend
+with room. `--list-devices` shows the heap:
+
+```
+ROCm0:   AMD Radeon Graphics (98304 MiB)
+Vulkan0: AMD Radeon Graphics (RADV STRIX_HALO) (113910 MiB)
+```
+
+## Memory Fit (Q6_K = 98.26 GiB)
+
+| Item | Size |
+| ---- | ---- |
+| Vulkan heap (llama.cpp view) | ~111.2 GiB |
+| Model Q6_K file | 98.26 GiB |
+| Model + MTP + graphs in heap | ~98.7 GiB |
+| Free heap after load | ~10 GiB |
+
+Ling uses gated MLA and KDA attention. KV cache stays tiny: 256k total
+context costs only ~2.5 GiB. This is why 256k context fits.
+
+## Files
+
+- `scripts/frank-large-ling.service` - systemd unit.
+- `scripts/serve-ling-flash.sh` - serve script (fork binary, Vulkan,
+  draft-mtp).
+- `scripts/frank-default-large.ling.env` - tunables.
+
+Passwordless management is installed:
+`/usr/local/sbin/frank-restart-ling.sh` and `/etc/sudoers.d/frank-ling`.
+
+## Switch Back to Qwen3.6-27B (MTP)
+
+The Qwen config was not changed. It remains committed as
+`frank-default-large.service` / `serve-llama-frank-large.sh` /
+`frank-default-large.env` (draft-mtp, n_max 3, ubatch 8192).
+
+Run these commands:
+
+```
+sudo systemctl stop frank-large-ling
+sudo systemctl start frank-default-large
+```
+
+## Caveats
+
+- The `bailingmoe3` fork validates CPU and CUDA upstream. Vulkan was
+  validated here (load + generation + concurrent streams) on 2026-08-14.
+- The tokenizer logs `special_eos_id is not in special_eog_ids` at load.
+  It is a warning; generation stops correctly at EOS.
+- Physical RAM pressure stays low because the model fits in the UMA
+  carve-out. Generation speed on this APU is below the Qwen3.6-27B MTP
+  config (~7.7 vs ~15-19 t/s at 4k).
+- The heap is effectively full (98.7 of 111.2 GiB). Do not raise the
+  context much without checking `--list-devices` free MiB.
+
+## See Also
+
+- [[frank-default-large-benchmark]]
+- [[DFlash DSpark Speculative Decoding]]
+- [[deployment]]