Skip to content
Language models

maintenance-work-order-triage-5M

An ultra-lightweight model for maintenance work order triage

2 min read

The model reads short equipment maintenance requests and instantly outputs two parameters: urgency (ranging from immediate shutdown to routine monitoring) and the specialized department (mechanics, electricians, hydraulics, and others). Built on a compact google/bert_uncased_L-4_H-128_A-2 encoder with two classification heads, it weighs just 18.28 MB with 4.78M parameters. Unlike the larger maintenance-work-order-triage-13M version, this model runs 3.6 times faster—about 1.8 ms per request on a standard CPU without a GPU or cloud connection. It is useful for manufacturing plants and engineers looking to automate primary ticket sorting on low-end hardware or edge devices.

Key features

  • High speed: Processes a single request on a CPU in 1.799 ms with no GPU required.
  • Two tasks at once: Determines the priority (4 classes) and the required maintenance department (6 classes), returning calibrated confidence for each output.
  • Two operating modes: A balanced mode and a safety mode that catches up to 84% of critical failures (STOP_NOW).
  • Smart escalation: Sends ambiguous or incomplete requests for manual review (ESCALATE) instead of guessing.

What you can do

  • Pre-sorting in asset management systems: Integrate as a local assistant for SAP PM, Maximo, or another ticketing system so planners only need to confirm pre-tagged requests.
  • Edge deployment: Run directly on a production terminal, mobile device, or low-power PC in a secure, air-gapped environment without internet access.
  • Emergency filtering: Enable safety mode to quickly highlight critical failures that demand immediate response.

Example

No code snippet card is included—here is an example of the model operating in balanced mode:
Input: "Hyd hose burst on press 3, oil spraying across walkway"
Result: priority STOP_NOW (confidence 0.97), department HYDRAULIC_PNEUMATIC (confidence 0.97).
For the request "Gearbox oil level low, top up", the model outputs PLAN (0.93) and LUBRICATION (0.98).

Read moreDetails
Tested on 1,500 real requests: macro-F1 for priority is 0.8265 (0.83), for department 0.8008 (0.80), and critical case recall is 64% in standard mode and 84% in safety mode.
Resources: 4.78M parameters, size 18.28 MB, latency 1.799 ms on CPU.
Limitations: English language only, short texts (mostly up to 8 words) describing a single malfunction; it does not replace human operators or industrial safety systems.
Training: Full fine-tuning on the MaintNorm and MaintIE datasets augmented with synthetic factory work orders.
Discuss Telegram 0 25

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 0.0 GB · ready files0.6 GB
  • BF16weights 0.0 GB0.6 GB
  • FP8 / INT8recommendedweights 0.0 GB0.6 GB
  • 4-bitweights 0.0 GB0.6 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · F32 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

A ready script for the machine we advise above: paste it and the model downloads and starts.

  1. 1. Rent A30

    $0.26 an hour with the coupon. Pick the Ubuntu image when you create the machine.

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # karthe05/maintenance-work-order-triage-5M — A30, FP8 / INT8
    # NVIDIA driver: CUDA 13 needs 580+. If older or missing, the current nvidia-open is installed (keep yours: KEEP_DRIVER=1)
    DRV=$(nvidia-smi --query-gpu=driver_version --format=csv,noheader 2>/dev/null | head -1 | cut -d. -f1)
    case "$DRV" in ""|*[!0-9]*) DRV=0 ;; esac
    if [ "$DRV" -lt 580 ] && [ -z "${KEEP_DRIVER:-}" ]; then
      . /etc/os-release && \
      curl -fsSLO "https://developer.download.nvidia.com/compute/cuda/repos/ubuntu${VERSION_ID//./}/x86_64/cuda-keyring_1.1-1_all.deb" && \
      sudo dpkg -i cuda-keyring_1.1-1_all.deb && rm -f cuda-keyring_1.1-1_all.deb && \
      sudo apt-get update && \
      sudo DEBIAN_FRONTEND=noninteractive apt-get install -y nvidia-open || \
      { echo "Could not install the NVIDIA driver: see the error above. To keep the current one, run with KEEP_DRIVER=1."; exit 1; }
      echo "NVIDIA driver updated. Reboot the machine (sudo reboot) and run this same script again."
      exit 0
    fi
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv venv ~/vllm --python 3.12 && source ~/vllm/bin/activate
    uv pip install vllm
    vllm serve karthe05/maintenance-work-order-triage-5M --max-model-len 8192 --quantization fp8
  3. 3. Use it

    The server listens on port 8000 and speaks the OpenAI API: http://localhost:8000/v1 on the machine. From your computer: ssh -L 8000:localhost:8000 user@machine-IP.

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

  • MTP (multi-token prediction)0.34.0from November 16, 2026new release

    1.5–2× generation speed

    If the model has MTP layers (DeepSeek-V3/R1, Qwen3-Next, GLM, MiMo and others), it predicts several tokens per step itself.

    How to enable: vLLM / SGLang: --speculative-config '{"method": "mtp", "num_speculative_tokens": 1}'. Check the model card.

    Earlier generations (2)
    • EAGLE-3 — Speculative decoding with a small draft head. For models without their own MTP layers.
    • EAGLE-2 / Medusa — Earlier draft-head generations. Still work if that is all you find.
  • FP8 KV cache0.34.0from November 16, 2026new release

    2× more context in the same memory

    Keeps the attention cache in 8 bits, so long contexts fit where they did not.

    How to enable: vLLM: --kv-cache-dtype fp8; llama.cpp: -ctk q8_0 -ctv q8_0.

  • vLLM / SGLang0.34.0from November 16, 2026new release

    many times more tokens/s on a server

    Server engines: batching, FlashAttention, FP8 and speculative decoding out of the box.

    How to enable: vllm serve <model> — an OpenAI-compatible API.

    Earlier generations (1)
    • llama.cpp — GGUF models on any GPU, Mac or even CPU; some layers can stay in RAM.

Model family

New posts about versions of bert_uncased_L-4_H-128_A-2 (quants, finetunes, LoRAs) will show up under the bell.

Fine-tunes 12All on Hugging Face

About the model

karthe05/maintenance-work-order-triage-5M
Author
karthe05
Category
Language models
Released
Posted

On NeyroLearn

Hugging Face

250

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments