Skip to content
Language models

Laya on Qualcomm NPU

binaries for fast decision-model inference on Qualcomm chips

2 min readupdated

This repository provides precompiled QNN context binaries (specialized context binary files for the Qualcomm AI Runtime engine) for the Laya decision-making model, targeting the Hexagon NPU on the Qualcomm QCS8550 chip in fp16 (half-precision floating-point) precision. The original model weights were converted with zero degradation in output accuracy—every decision matches the CPU baseline identically. This delivers a 15–39x speedup on embedded hardware compared to running on a standard device CPU, making it ideal for developers building embedded systems and edge AI applications on the Qualcomm platform.

Key features

  • Massive speedup: Inference runs 15–39x faster depending on context length and checkout.
  • Exact accuracy: Matches CPU baseline outputs across 48/48 tests with zero loss in decision quality.
  • Ready-to-use binaries: Includes three precompiled variants covering the English, multilingual, and typed-decision models.
  • Sustained efficiency: Maintains steady throughput over five-minute runs without throttling or latency degradation.

What you can do

  • Deploy to edge hardware: Run the decision-making model directly on Qualcomm QCS8550 devices such as the Thundercomm RB5 Gen2.
  • Run via Python: Use ready-made classes from the companion repository for fast text classification and task routing.
  • Customize builds: Use the open-source export and compilation scripts to adapt the model to custom requirements or target environments.

Example

Sample on-device inference using the dedicated router from the open-source repository:

from laya_npu import NPURouter            # from the code repository
router = NPURouter()                      # one NPUAgent per models/.bin
print(router.predict("We were billed twice for March. Refund it today or we cancel.", {
    "department": {"type": "choice", "instructions": "Which department should handle this?",
                   "criteria": {"billing": "invoices, payments, refunds",
                                "technical": "bugs, outages, errors", "other": "everything else"}},
    "churn_risk": {"type": "noul", "instructions": "Does the customer threaten to leave?"}}))
Read moreDetails
Benchmarks: At a 128-token context length, the English variant runs in 49 ms (a 27x speedup over CPU); at 512 tokens, it completes in 148 ms (a 31x speedup).
Hardware requirements: Targets the Qualcomm QCS8550 SoC (Hexagon HTP V73) and requires QAIRT 2.51 or newer.
Limitations: Maximum batch size is 1, choice count is capped at 32 options per query, and binaries are tightly coupled to the specific SDK version and target chip.
License: Apache-2.0 (inherited from the original Laya model).
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.8 GB · ready files1.5 GB
  • FP8 / INT8recommendedweights 0.4 GB1.0 GB
  • 4-bitweights 0.2 GB0.9 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 · BF16 — 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.

    # kadamiak-consultred/laya-qcs8550-npu — 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 kadamiak-consultred/laya-qcs8550-npu --max-model-len 8192 --quantization fp8
    Add more LoRAs to laya-qcs8550-npu
  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 laya (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 1All on Hugging Face

About the model

kadamiak-consultred/laya-qcs8550-npu
Author
kadamiak-consultred
Category
Language models
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

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

Comments