LM

Topics

Agent	Model	Runner	Driver
axe SKILL MCP	Management	SGLang llama.cpp for SYCL llama.cpp for OpenVINO	xe Driver oneAPI OpenVINO

Decision Tree for tinkering with Intel Arc Pro B70

File Path	/images/quickmmd/LM-tinkering-b70.svg
File Size	47.97 KB
Last modified	21:54:49
Time of SVG loading	2.31 second(s)
Time of SVG convertion	0.00 second(s)
Cache Detected	No
MD5 of incoming syntax	6c1827bb6f6a512a8e4ffcb71e384f36
MD5 of cached syntax	47d19aaedfbb9b5eb98f18ddfd23af9d
SVG Engine	Kroki API
Kroki API URL	http://kroki:8000
Mermaid syntax Reference	https://mermaid.js.org/intro/syntax-reference.html
QuickMMD Version	1.0.0 - About QuickMMD

flowchart LR
  subgraph hardware
    A[Got B70]
  end

  subgraph os
    B1[Install<br>Ubuntu 24.04]:::node-okay
    B2[Install<br>Ubuntu 26.04]:::node-warn
    A --> B1
    A --> B2
  end

  subgraph drvier
    C1[Install<br>PPA driver]:::node-okay
    C2[Install<br>oneAPI]:::node-okay
    C3[Install<br>OpenVINO]:::node-okay
    C4[Install<br>PPA driver]:::node-warn
    C5[Cannot install<br>oneAPI]:::node-fail
    C6[Cannot install<br>OpenVINO]:::node-fail
    C1 --> C2 & C3
    C4 --> C5 & C6
    B1 --> C1
    B2 --> C4
  end

  subgraph runtime
    D1[Build llama.cpp for<br>SYCL]:::node-okay
    D2[Build llama.cpp for<br>OpenVINO]:::node-okay
    D3[Build SGLang for<br>SYCL]:::node-todo
    D4[Build SGLang for<br>OpenVINO]:::node-todo
    C2 --> D1
    C3 --> D2
    C2 --> D3
    C3 --> D4
  end

  subgraph tools
    T1[Install<br>pipx]
    T2[Install<br>hf]
    T3[Install<br>nvtop]
    T4[Install<br>screen]
    T5[Install<br>llama-swap]
    T1 --> T2
    B1 --> T1
    B1 --> T3
    B1 --> T4
    B1 --> T5
  end

  subgraph models
    M1[Muse-Glimmer-30B]
    M2[Devstral-Small-2]
    M3[Qwen3-Coder-30B]
    M4[Gemma-4-E4B]
    T2 --> M1 & M2 & M3 & M4
  end

  subgraph env
    V1[Optimize<br>.bashrc]
    V2[Optimize<br>llama-swap.yaml]
    C2 --> V1
    C3 --> V1
    T2 --> V1
    T4 ---> V1
    T5 ---> V2
  end

  subgraph service
    Z[service]
    M1 --> Z
    D1 --> Z
    V2 --> Z
  end

  classDef node-okay fill:#efe,stroke:#393
  classDef node-warn fill:#fff0e0,stroke:#d63
  classDef node-fail fill:#fee,stroke:#d33
  classDef node-todo fill:#eee,stroke:#777

Build environment

hf (model management)

• https://huggingface.co/docs/huggingface_hub/package_reference/environment_variables
• https://www.datalearner.com/en/leaderboards/category/code?benchmark=SWE-bench+Verified&modelSize=34b&licenseType=open
• https://huggingface.co/datasets/ScaleAI/SWE-bench_Pro

Purpose	Command
cache management	hf cache list hf cache rm <model id> hf cache prune
fix WiFi problem	HF_XET_FIXED_DOWNLOAD_CONCURRENCY=10 hf download "unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF" --include "*UD-Q4_K_XL*" HF_XET_FIXED_DOWNLOAD_CONCURRENCY=10 hf download "unsloth/Devstral-Small-2-24B-Instruct-2512-GGUF" --include "*UD-Q4_K_XL*"
search	# search by population hf models ls --search "Muse-Glimmer" --apps llama.cpp --expand "downloads,likes,createdAt,lastModified" --sort downloads --no-truncate --limit 25 # search for latest publish hf models ls --search "Muse-Glimmer" --apps llama.cpp --expand "downloads,likes,createdAt,lastModified" --sort created_at --no-truncate --limit 25 # search for latest tunning hf models ls --search "Muse-Glimmer" --apps llama.cpp --expand "downloads,likes,createdAt,lastModified" --sort last_modified --no-truncate --limit 25
optimize Qwen3-Coder	hf models ls --search "coder" --apps llama.cpp --sort downloads --limit 1 --format json | jq . hf models ls -h "unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF" hf download "unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF" --include "*UD-Q4_K_XL*"
optimize gemma-4-E4B	hf models ls -h unsloth/gemma-4-E4B-it-qat-GGUF hf download unsloth/gemma-4-E4B-it-qat-GGUF --include "*UD-Q4_K_XL*" hf download unsloth/gemma-4-E4B-it-qat-GGUF --include "mmproj-BF16.gguf" hf download unsloth/gemma-4-E4B-it-qat-GGUF --include "mtp-gemma-4-E4B-it.gguf"
optimize gemma-4-12B	hf models ls -h unsloth/gemma-4-12B-it-qat-GGUF hf download unsloth/gemma-4-12B-it-qat-GGUF --include "*UD-Q4_K_XL*" hf download unsloth/gemma-4-12B-it-qat-GGUF --include "mmproj-BF16.gguf" hf download unsloth/gemma-4-12B-it-qat-GGUF --include "mtp-gemma-4-12B-it.gguf"
optimize Muse Glimmer	hf models ls -h unsloth/Muse-Glimmer-30B-GGUF hf download unsloth/Muse-Glimmer-30B-GGUF --include "*UD-Q4_K_XL*" hf download unsloth/Muse-Glimmer-30B-GGUF --include "*UD-Q5_K_XL*" hf download unsloth/Muse-Glimmer-30B-GGUF --include "*UD-Q6_K_XL*" hf download unsloth/Muse-Glimmer-30B-GGUF --include "dflash-kquant.gguf" hf download unsloth/Muse-Glimmer-30B-GGUF --include "mmproj-kquant.gguf"
	tree ~/.cache/huggingface/hub/models--unsloth--Devstral-Small-2-24B-Instruct-2512-GGUF/snapshots tree ~/.cache/huggingface/hub/models--unsloth--Qwen3-Coder-30B-A3B-Instruct-GGUF/snapshots

Ubuntu

• LLM/Intel Pro Arc B70
• LLM/Ryzen AI 7 350 (XDNA)

Agents

• LLM/jrswab_axe

Wishlist

• LLM/prompt for coding