Skip to content
Other

Chess league bot NLU

компактный NLU для шахматного бота

2 min readupdated

Это компактная GLiNER2-модель на 74M параметров (0,07B), дообученная от fastino/gliner2.5-small-v1 для arboters — бота лиги Lichess4545 в Zulip. Она определяет команды в сообщениях и извлекает из переписки предложенные, принятые и отозванные игровые слоты. Обучена 5 эпох на синтетическом датасете, работает с английским.

Ключевые особенности

  • 74M параметров, GLiNER2; на CPU один forward pass — около 50–80 мс через ONNX Runtime, до ~190 мс на длинных тредах.
  • Команды: классификация command и event плюс извлечение span'ов аргументов; топ-1 команда — ответ модели.
  • Планирование: находит offered time, accepted time и withdrawn time; согласие/отзыв привязываются к исходному предложению.
  • На тесте: effort verdict 98%, offered time F1 0.99, accepted 0.94, withdrawn 0.92, command accuracy 80%.
  • Только синтетические данные; отказы могут читаться как согласие, команды часто уходят в none, recipient не выучен.

Пример

from gliner2 import AutoExtractor
model = AutoExtractor.from_pretrained('Lichess4545/chess-league-bot-nlu', map_location='cpu')
schema = model.create_schema().entities(labels['scheduling'])
model.extract('white: can we do thu 19:00 or sat 14:00?\nblack: sat works\n', schema)
ПодробнееГлавные метрики end-to-end: effort 98%, offered time F1 0.99, accepted 0.94, withdrawn 0.92; command accuracy 80% (78% на сообщениях, которых не было в обучении).
Инференс: ~50–80 мс на одно ядро CPU с ONNX Runtime, до ~190 мс на самых длинных тредах.
Ограничения: обучение только на синтетике, реальные сообщения будут сложнее; отказы иногда принимаются за согласие, слабые команды — available, tell, assign_alternate, pairing, help; recipient не выучен.
Лицензия AGPL-3.0-or-later; базовая модель — Apache 2.0.
Discuss Telegram 1 30

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 0.5 GB · ready files1.1 GB
  • BF16weights 0.3 GB0.8 GB
  • FP8 / INT8recommendedweights 0.1 GB0.6 GB
  • 4-bitweights 0.1 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

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    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.

    # Lichess4545/chess-league-bot-nlu
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download Lichess4545/chess-league-bot-nlu --local-dir ~/models/chess-league-bot-nlu
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to chess-league-bot-nlu
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

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.

Model family

New posts about versions of SmolLM2-135M (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 2All on Hugging Face

About the model

Lichess4545/chess-league-bot-nlu
Author
Lichess4545
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

301

Ask AI about this model

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

Comments