Frank 0e5c37f2f8 frank-default-small: use q8_0 (not q8_KV) KV cache on ik_llama ...

The ik_llama fast q8_KV cache type crashes at model init on this
QWEN3MOE/MTP + multimodal arch at commit b166e269 with a ggml_view size
assert (ggml.c:5427) right after the output buffer is created, looping
the service. Empirically isolated by toggling cache type: q8_0 and f16
load and serve (smoke-tested correct output), q8_KV asserts. Switch to
q8_0 + Hadamard (K/V), which still gives the 8-bit cache + Hadamard
quality win. KB updated with the crash lesson.

2026-08-26 03:23:18 -07:00

..

Pipeline Parallelism

Doc audit: fix all broken links + correct stale facts across the KB

2026-08-24 01:10:44 -07:00

ciru-activefpx-promptforge.md

Update benchmark documentation after harness replacement

2026-08-25 19:08:42 -07:00

Dankclaw Improvements.md

Upgrade frank-default-large to kingjones Q8_0 AGENT at full 262144 context

2026-08-25 07:12:02 -07:00

Dankclaw System Specs.md

Swap LLM fleet to Ornith CRACK (small) and kingjones ROCmFP4 STRIX MTP (large); apply Qwen-Sharp chat template

2026-08-24 21:35:11 -07:00

Dankserver System Specs.md

Use ngram-mod and prioritize attention experts

2026-08-25 01:40:06 -07:00

dankserver-vram-growth.md

Reorganize: llm/docs -> docs/LLM (LLM fleet knowledge base)

2026-08-22 15:06:51 -07:00

dealign-ornith-crack.md

Use ngram-mod and prioritize attention experts

2026-08-25 01:40:06 -07:00

deployment.md

Deploy services independently by changed paths

2026-08-25 20:13:33 -07:00

dflash2-large-vulkan.md

Reorganize: llm/docs -> docs/LLM (LLM fleet knowledge base)

2026-08-22 15:06:51 -07:00

DFlash DSpark Speculative Decoding.md

Doc audit: fix all broken links + correct stale facts across the KB

2026-08-24 01:10:44 -07:00

eemin-rocmi4-mtp.md

Update benchmark documentation after harness replacement

2026-08-25 19:08:42 -07:00

FastFlowLM NPU Inference.md

Reorganize: llm/docs -> docs/LLM (LLM fleet knowledge base)

2026-08-22 15:06:51 -07:00

frank-default-large-benchmark.md

Reorganize: llm/docs -> docs/LLM (LLM fleet knowledge base)

2026-08-22 15:06:51 -07:00

frank-default-small-benchmark.md

Record no-spec ubatch benchmark

2026-08-25 01:29:49 -07:00

frank-redo-2026-08-17.md

Reorganize: llm/docs -> docs/LLM (LLM fleet knowledge base)

2026-08-22 15:06:51 -07:00

Hybrid CPU GPU Processing Dankserver.md

Reorganize: llm/docs -> docs/LLM (LLM fleet knowledge base)

2026-08-22 15:06:51 -07:00

ikllama-ornith-dankserver.md

frank-default-small: use q8_0 (not q8_KV) KV cache on ik_llama

2026-08-26 03:23:18 -07:00

index.md

frank-default-small: use q8_0 (not q8_KV) KV cache on ik_llama

2026-08-26 03:23:18 -07:00

kairic-edge.md

Reorganize: llm/docs -> docs/LLM (LLM fleet knowledge base)

2026-08-22 15:06:51 -07:00

kingjones-q8agent-rocmfpx.md

Clarify historical ROCmFPX payload sizes

2026-08-25 13:28:38 -07:00

kingjones-strix-mtp.md

Fix Q6 AGENT projector path

2026-08-25 13:58:50 -07:00

ling-3.0-flash-trial.md

Reorganize: llm/docs -> docs/LLM (LLM fleet knowledge base)

2026-08-22 15:06:51 -07:00

ling-3.0-flash-vllm-plan.md

Reorganize: llm/docs -> docs/LLM (LLM fleet knowledge base)

2026-08-22 15:06:51 -07:00

Quantization by Hardware Architecture.md

Reorganize: llm/docs -> docs/LLM (LLM fleet knowledge base)

2026-08-22 15:06:51 -07:00

qwen3.8-spec-matrix.md

Reorganize: llm/docs -> docs/LLM (LLM fleet knowledge base)

2026-08-22 15:06:51 -07:00

qwen3.8-strix-optimizations.md

Upgrade frank-default-large to kingjones Q8_0 AGENT at full 262144 context

2026-08-25 07:12:02 -07:00

RAM CPU Offloading in VLLM.md

Reorganize: llm/docs -> docs/LLM (LLM fleet knowledge base)

2026-08-22 15:06:51 -07:00

Running Qwen3.6 27B FP8 on Dankclaw.md

Reorganize: llm/docs -> docs/LLM (LLM fleet knowledge base)

2026-08-22 15:06:51 -07:00

Running Qwen3.6 35B A3B FP8 on Dankserver.md

Reorganize: llm/docs -> docs/LLM (LLM fleet knowledge base)

2026-08-22 15:06:51 -07:00

Shared Memory Architecture Dankclaw.md

Reorganize: llm/docs -> docs/LLM (LLM fleet knowledge base)

2026-08-22 15:06:51 -07:00

ubatch-benchmark-large.md

Set equal batch and ubatch at 8k context

2026-08-25 21:04:00 -07:00

VLLM Index.md

Doc audit: fix all broken links + correct stale facts across the KB

2026-08-24 01:10:44 -07:00

VLLM Overview.md

Reorganize: llm/docs -> docs/LLM (LLM fleet knowledge base)

2026-08-22 15:06:51 -07:00

vllm_research_findings.md

Reorganize: llm/docs -> docs/LLM (LLM fleet knowledge base)

2026-08-22 15:06:51 -07:00