# pixelml/club-170hx architecture

> Architecture diagram of the GitHub repository pixelml/club-170hx, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-01). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/pixelml/club-170hx
- Repository: https://github.com/pixelml/club-170hx

## Overview

Club CMP 170HX is a community-tested collection of hardware diagnostics, AI workload recipes, and reproducible benchmark experiments—not a single application. A typical operator runs a model recipe against a live serving endpoint, checks runtime and card telemetry, records usage-accounted measurements, and validates results. The repository also preserves independent long-context and multimodal benchmarks, Jev-compatible classification experiments, and model-specific vLLM patches. The tree and excerpts establish only selected connections; unsampled experiment families are represented by their directories, without inferred call chains.

Main entry points: [`scripts/build_qwen_recipe_notebook.py`](https://github.com/pixelml/club-170hx/blob/main/scripts/build_qwen_recipe_notebook.py) (Recipe runner), [`results/2026-09-20-qwen3.8-27b-w4a16-jev-1card-vllm/harness/jev_probe.py`](https://github.com/pixelml/club-170hx/blob/main/results/2026-09-20-qwen3.8-27b-w4a16-jev-1card-vllm/harness/jev_probe.py) (Jev probes), [`results/2026-09-22-mimo-v2.6-flash-4card-pp4-vllm/patches/mimo_v2.py`](https://github.com/pixelml/club-170hx/blob/main/results/2026-09-22-mimo-v2.6-flash-4card-pp4-vllm/patches/mimo_v2.py) (MiMo model), [`recipes/qwen3.8-27b-dflash2/live_benchmark.py`](https://github.com/pixelml/club-170hx/blob/main/recipes/qwen3.8-27b-dflash2/live_benchmark.py) (Usage benchmark)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_recipe["Live recipe"]
  node_recipe_builder["Recipe runner"]
  node_live_bench["Usage benchmark<br/>[live_benchmark.py]"]
  node_live_support["Runtime safety<br/>[live_support.py]"]
  node_pinned_models["Model preparation"]
  node_result_pipeline["Result validation<br/>[result_pipeline.py]"]
end

subgraph group_experiments["Benchmark experiments"]
  node_jev_server["Jev classifier<br/>[jev_server.py]"]
  node_jev_probe["Jev probes<br/>[jev_probe.py]"]
  node_jev_harness["Jev measurements<br/>[harness]"]
  node_longctx_driver["Long-context driver<br/>[run_longctx.py]"]
  node_longctx_bench["Long-context tests<br/>[longctx_bench.py]"]
  node_experiment_suite["Other benchmarks"]
  node_notebook_experiments["Executed experiments"]
end

subgraph group_model["Model extensions"]
  node_mimo_core["MiMo model<br/>[mimo_v2.py]"]
  node_mimo_mtp["MiMo predictor<br/>[mimo_v2_mtp.py]"]
  node_mimo_omni["MiMo vision<br/>[mimo_v2_omni.py]"]
  node_other_patches["Other model patches<br/>[patches]"]
end

subgraph group_diagnostics["Hardware diagnostics"]
  node_qc_checks["Card diagnostics<br/>[memtest-select.py]"]
  node_health_gate["Compute health gate<br/>[matmul_check.py]"]
  node_power_tests["Power experiments"]
  node_cost_analysis["Cost analysis<br/>[make_chart.py]"]
end

node_operator(("Operator"))
node_serving_endpoint["Serving endpoint"]

node_operator -->|"runs"| node_recipe_builder
node_recipe_builder -->|"uses source"| node_live_bench
node_recipe_builder -->|"uses source"| node_live_support
node_recipe_builder -->|"uses source"| node_result_pipeline
node_live_bench -->|"sends requests"| node_serving_endpoint
node_operator -->|"runs probes"| node_jev_probe
node_jev_probe -->|"calls endpoint"| node_jev_server
node_jev_probe -.->|"queries upstream"| node_serving_endpoint
node_operator -->|"runs tests"| node_longctx_driver
node_longctx_driver -->|"calls tests"| node_longctx_bench
node_operator -.->|"runs experiments"| node_experiment_suite
node_operator -.->|"reproduces results"| node_notebook_experiments
node_operator -.->|"checks card"| node_qc_checks
node_operator -.->|"checks compute"| node_health_gate
node_operator -.->|"measures power"| node_power_tests
node_operator -.->|"compares cost"| node_cost_analysis
node_mimo_mtp -->|"imports model"| node_mimo_core
node_mimo_omni -->|"imports model"| node_mimo_core
node_operator -.->|"uses patch"| node_mimo_core
node_operator -.->|"uses patches"| node_other_patches

click node_recipe_builder "https://github.com/pixelml/club-170hx/blob/main/scripts/build_qwen_recipe_notebook.py"
click node_live_bench "https://github.com/pixelml/club-170hx/blob/main/recipes/qwen3.8-27b-dflash2/live_benchmark.py"
click node_live_support "https://github.com/pixelml/club-170hx/blob/main/recipes/qwen3.8-27b-dflash2/live_support.py"
click node_pinned_models "https://github.com/pixelml/club-170hx/blob/main/recipes/qwen3.8-27b-dflash2/prepare_pinned_models.py"
click node_result_pipeline "https://github.com/pixelml/club-170hx/blob/main/recipes/qwen3.8-27b-dflash2/result_pipeline.py"
click node_jev_server "https://github.com/pixelml/club-170hx/blob/main/results/2026-09-20-qwen3.8-27b-w4a16-jev-1card-vllm/harness/jev_server.py"
click node_jev_probe "https://github.com/pixelml/club-170hx/blob/main/results/2026-09-20-qwen3.8-27b-w4a16-jev-1card-vllm/harness/jev_probe.py"
click node_jev_harness "https://github.com/pixelml/club-170hx/tree/main/results/2026-09-20-qwen3.8-27b-w4a16-jev-1card-vllm/harness"
click node_longctx_driver "https://github.com/pixelml/club-170hx/blob/main/results/2026-09-02-deepseek-v4-flash-vision-exp-4card-longctx-262k/run_longctx.py"
click node_longctx_bench "https://github.com/pixelml/club-170hx/blob/main/results/2026-09-02-deepseek-v4-flash-vision-exp-4card-longctx-262k/longctx_bench.py"
click node_experiment_suite "https://github.com/pixelml/club-170hx/tree/main/results"
click node_notebook_experiments "https://github.com/pixelml/club-170hx/tree/main/notebooks"
click node_mimo_core "https://github.com/pixelml/club-170hx/blob/main/results/2026-09-22-mimo-v2.6-flash-4card-pp4-vllm/patches/mimo_v2.py"
click node_mimo_mtp "https://github.com/pixelml/club-170hx/blob/main/results/2026-09-22-mimo-v2.6-flash-4card-pp4-vllm/patches/mimo_v2_mtp.py"
click node_mimo_omni "https://github.com/pixelml/club-170hx/blob/main/results/2026-09-22-mimo-v2.6-flash-4card-pp4-vllm/patches/mimo_v2_omni.py"
click node_other_patches "https://github.com/pixelml/club-170hx/tree/main/results/2026-09-22-mimo-v2.6-flash-4card-pp4-vllm/patches"
click node_qc_checks "https://github.com/pixelml/club-170hx/blob/main/scripts/qc/memtest-select.py"
click node_health_gate "https://github.com/pixelml/club-170hx/blob/main/results/2026-09-02-health-gate/receipts/matmul_check.py"
click node_power_tests "https://github.com/pixelml/club-170hx/blob/main/results/2026-09-02-deepseek-v4-flash-vision-exp-4card-repro-power/bench_repro_power.py"
click node_cost_analysis "https://github.com/pixelml/club-170hx/blob/main/results/2026-09-02-cross-platform-cost-watt/make_chart.py"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_recipe_builder,node_live_bench,node_live_support,node_pinned_models,node_result_pipeline toneBlue
class node_jev_server,node_jev_probe,node_jev_harness,node_longctx_driver,node_longctx_bench,node_experiment_suite,node_notebook_experiments toneAmber
class node_mimo_core,node_mimo_mtp,node_mimo_omni,node_other_patches toneMint
class node_qc_checks,node_health_gate,node_power_tests,node_cost_analysis toneRose
class node_operator,node_serving_endpoint toneIndigo
```

## Components

### Live recipe

- [Recipe runner](https://github.com/pixelml/club-170hx/blob/main/scripts/build_qwen_recipe_notebook.py): `scripts/build_qwen_recipe_notebook.py`
- [Usage benchmark](https://github.com/pixelml/club-170hx/blob/main/recipes/qwen3.8-27b-dflash2/live_benchmark.py): `recipes/qwen3.8-27b-dflash2/live_benchmark.py`
- [Runtime safety](https://github.com/pixelml/club-170hx/blob/main/recipes/qwen3.8-27b-dflash2/live_support.py): `recipes/qwen3.8-27b-dflash2/live_support.py`
- [Model preparation](https://github.com/pixelml/club-170hx/blob/main/recipes/qwen3.8-27b-dflash2/prepare_pinned_models.py): `recipes/qwen3.8-27b-dflash2/prepare_pinned_models.py`
- [Result validation](https://github.com/pixelml/club-170hx/blob/main/recipes/qwen3.8-27b-dflash2/result_pipeline.py): `recipes/qwen3.8-27b-dflash2/result_pipeline.py`

### Benchmark experiments

- [Jev classifier](https://github.com/pixelml/club-170hx/blob/main/results/2026-09-20-qwen3.8-27b-w4a16-jev-1card-vllm/harness/jev_server.py): `results/2026-09-20-qwen3.8-27b-w4a16-jev-1card-vllm/harness/jev_server.py`
- [Jev probes](https://github.com/pixelml/club-170hx/blob/main/results/2026-09-20-qwen3.8-27b-w4a16-jev-1card-vllm/harness/jev_probe.py): `results/2026-09-20-qwen3.8-27b-w4a16-jev-1card-vllm/harness/jev_probe.py`
- [Jev measurements](https://github.com/pixelml/club-170hx/tree/main/results/2026-09-20-qwen3.8-27b-w4a16-jev-1card-vllm/harness): `results/2026-09-20-qwen3.8-27b-w4a16-jev-1card-vllm/harness`
- [Long-context driver](https://github.com/pixelml/club-170hx/blob/main/results/2026-09-02-deepseek-v4-flash-vision-exp-4card-longctx-262k/run_longctx.py): `results/2026-09-02-deepseek-v4-flash-vision-exp-4card-longctx-262k/run_longctx.py`
- [Long-context tests](https://github.com/pixelml/club-170hx/blob/main/results/2026-09-02-deepseek-v4-flash-vision-exp-4card-longctx-262k/longctx_bench.py): `results/2026-09-02-deepseek-v4-flash-vision-exp-4card-longctx-262k/longctx_bench.py`
- [Other benchmarks](https://github.com/pixelml/club-170hx/tree/main/results): `results`
- [Executed experiments](https://github.com/pixelml/club-170hx/tree/main/notebooks): `notebooks`

### Model extensions

- [MiMo model](https://github.com/pixelml/club-170hx/blob/main/results/2026-09-22-mimo-v2.6-flash-4card-pp4-vllm/patches/mimo_v2.py): `results/2026-09-22-mimo-v2.6-flash-4card-pp4-vllm/patches/mimo_v2.py`
- [MiMo predictor](https://github.com/pixelml/club-170hx/blob/main/results/2026-09-22-mimo-v2.6-flash-4card-pp4-vllm/patches/mimo_v2_mtp.py): `results/2026-09-22-mimo-v2.6-flash-4card-pp4-vllm/patches/mimo_v2_mtp.py`
- [MiMo vision](https://github.com/pixelml/club-170hx/blob/main/results/2026-09-22-mimo-v2.6-flash-4card-pp4-vllm/patches/mimo_v2_omni.py): `results/2026-09-22-mimo-v2.6-flash-4card-pp4-vllm/patches/mimo_v2_omni.py`
- [Other model patches](https://github.com/pixelml/club-170hx/tree/main/results/2026-09-22-mimo-v2.6-flash-4card-pp4-vllm/patches): `results/2026-09-22-mimo-v2.6-flash-4card-pp4-vllm/patches`

### Hardware diagnostics

- [Card diagnostics](https://github.com/pixelml/club-170hx/blob/main/scripts/qc/memtest-select.py): `scripts/qc/memtest-select.py`
- [Compute health gate](https://github.com/pixelml/club-170hx/blob/main/results/2026-09-02-health-gate/receipts/matmul_check.py): `results/2026-09-02-health-gate/receipts/matmul_check.py`
- [Power experiments](https://github.com/pixelml/club-170hx/blob/main/results/2026-09-02-deepseek-v4-flash-vision-exp-4card-repro-power/bench_repro_power.py): `results/2026-09-02-deepseek-v4-flash-vision-exp-4card-repro-power/bench_repro_power.py`
- [Cost analysis](https://github.com/pixelml/club-170hx/blob/main/results/2026-09-02-cross-platform-cost-watt/make_chart.py): `results/2026-09-02-cross-platform-cost-watt/make_chart.py`

### Other components

- **Operator**
- **Serving endpoint**

## Connections

- Operator → Recipe runner: runs (evidence: `scripts/build_qwen_recipe_notebook.py`)
- Recipe runner → Usage benchmark: uses source (evidence: `scripts/build_qwen_recipe_notebook.py`)
- Recipe runner → Runtime safety: uses source (evidence: `scripts/build_qwen_recipe_notebook.py`)
- Recipe runner → Result validation: uses source (evidence: `scripts/build_qwen_recipe_notebook.py`)
- Usage benchmark → Serving endpoint: sends requests (evidence: `recipes/qwen3.8-27b-dflash2/live_benchmark.py`)
- Operator → Jev probes: runs probes (evidence: `results/2026-09-20-qwen3.8-27b-w4a16-jev-1card-vllm/harness/jev_probe.py`)
- Jev probes → Jev classifier: calls endpoint (evidence: `results/2026-09-20-qwen3.8-27b-w4a16-jev-1card-vllm/harness/jev_probe.py`)
- Jev probes → Serving endpoint: queries upstream
- Operator → Long-context driver: runs tests (evidence: `results/2026-09-02-deepseek-v4-flash-vision-exp-4card-longctx-262k/run_longctx.py`)
- Long-context driver → Long-context tests: calls tests (evidence: `results/2026-09-02-deepseek-v4-flash-vision-exp-4card-longctx-262k/run_longctx.py`)
- Operator → Other benchmarks: runs experiments
- Operator → Executed experiments: reproduces results
- Operator → Card diagnostics: checks card
- Operator → Compute health gate: checks compute
- Operator → Power experiments: measures power
- Operator → Cost analysis: compares cost
- MiMo predictor → MiMo model: imports model (evidence: `results/2026-09-22-mimo-v2.6-flash-4card-pp4-vllm/patches/mimo_v2_mtp.py`)
- MiMo vision → MiMo model: imports model (evidence: `results/2026-09-22-mimo-v2.6-flash-4card-pp4-vllm/patches/mimo_v2_omni.py`)
- Operator → MiMo model: uses patch
- Operator → Other model patches: uses patches
