Skip to content
Other

Dante Qwen2.5-0.5B LoRA

стиль «Божественной комедии» за 5 минут на T4

2 min read

Это LoRA-адаптер для Qwen/Qwen2.5-0.5B, который переносит модель в стиль «La Divina Commedia» Данте Алигьери. Обучен на итальянском тексте из Project Gutenberg eBook #1000 и обновляет около 1.8% параметров базовой модели.

Ключевые особенности

  • База — Qwen/Qwen2.5-0.5B; адаптер нацелен на итальянский язык и стиль «Комедии».
  • LoRA: r=16, alpha=32, dropout 0.05; обновляются q/k/v/o и gate/up/down-проекции, ~1.8% параметров.
  • Обучение на бесплатном Google Colab T4 заняло около 5 минут: 3 эпохи, 256-токенные блоки, effective batch 16, lr 2e-4 cosine, fp16.
  • Оценка в bits per character на одном валидационном тексте: 1.791 у адаптера, 2.220 у базовой Qwen2.5-0.5B и 2.005 у DanteGPT 10.65M from scratch.
  • Есть родственная модель omarkilbi/dante-gpt-scratch — 10.65M GPT, обученная с нуля на том же тексте.

Пример

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
tok = AutoTokenizer.from_pretrained('Qwen/Qwen2.5-0.5B')
base = AutoModelForCausalLM.from_pretrained('Qwen/Qwen2.5-0.5B')
model = PeftModel.from_pretrained(base, 'omarkilbi/dante-qwen2.5-0.5b-lora')
inp = tok('O tu che paghi il pedaggio', return_tensors='pt')
out = model.generate(**inp, max_new_tokens=150, temperature=0.8, top_p=0.95)
print(tok.decode(out[0], skip_special_tokens=True))
ПодробнееБенчмарк: validation bits per character — 1.791 у LoRA, 2.220 у базовой Qwen2.5-0.5B, 2.005 у DanteGPT 10.65M.
Железо: бесплатный Google Colab T4, обучение ~5 минут; обновлено ~1.8% параметров.
Ограничения: учебный проект, не для литературных или фактических задач; валидационные цифры частично затронуты memorization «Комедии»; модель иногда выдумывает слова и выдаёт заголовки вроде «Purgatorio / Canto XXIV».
Discuss Telegram 1 26

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 1.0 GB · ready files1.5 GB
  • FP8 / INT8recommendedweights 0.5 GB1.0 GB
  • 4-bitweights 0.3 GB0.8 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.

This is a LoRA: counted together with its base model Qwen2.5-0.5B.

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.

    # omarkilbi/dante-qwen2.5-0.5b-lora
    # ===== Tokens: paste yours between the quotes =====
    export HF_TOKEN=""        # huggingface.co/settings/tokens, for gated models
    # ==========================================
    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 Qwen/Qwen2.5-0.5B --local-dir ~/models/Qwen2.5-0.5B
    hf download omarkilbi/dante-qwen2.5-0.5b-lora --local-dir ~/models/dante-qwen2.5-0.5b-lora
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to Qwen2.5-0.5B
  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

Requires the base model

Qwen/Qwen2.5-0.5B

This model is a LoRA / adapter. Download the base model first, then load this one on top.

New posts about versions of Qwen2.5-0.5B (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 5All on Hugging Face
Merges 2All on Hugging Face

About the model

omarkilbi/dante-qwen2.5-0.5b-lora
Author
omarkilbi
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

261

Ask AI about this model

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

Comments

Language models

Carbon-X

Carbon-X
Language models

fikirmatik

fikirmatik