/ Модель

Характеристики

Контекст
32 768 токенов
Параметры
14 млрд
Размер весов
29,5 ГБ
Лицензия
Apache 2.0
Архитектура
qwen2 · dense
Базовая модель
Qwen2.5-14B-Instruct
Библиотека
transformers
Опубликована
20 сен 2024

Qwen2.5-Lumen-14B

Генерация текстаТекст → текст3 профессии

Qwen2.5-Lumen-14B — объединённая модель на базе Qwen2.5-14B-Instruct для следования инструкциям, художественного письма и ролевых диалогов.

138скачиваний за 30 дней21лайквеса обновлены3 окт 2024

Бенчмарки

Как модель справилась с задачами пользователей ailist и что о ней публикуют открытые лидерборды

Тесты пользователейпока нет

Никто ещё не опубликовал тест этой модели. Соберите бенчмарк из своих задач — результат появится здесь.

  1. Добавьте модель в бенчмарк

    Вместе с другими моделями категории «Текст» — сравнение идёт внутри одной категории.

  2. Опишите свои задачи

    Напишите запросы, при необходимости приложите файлы и эталонные ответы.

  3. Запустите и опубликуйте

    Модели ответят на каждую задачу, судья оценит ответы. Опубликованный результат появится на этой странице.

Публичные оценки6

Из открытых лидербордов. Сравнивайте результаты одного теста при одинаковых условиях — отсутствие оценки не значит, что модель хуже.

  • GPQA

    GPQA Raw · выше — лучше

    Результат

    0,328

    Open LLM v2 · архив

    Публичная таблицадругая ревизия

  • MATH Lvl 5

    MATH Lvl 5 Raw · выше — лучше

    Результат

    0,5363

    Open LLM v2 · архив

    Публичная таблицадругая ревизия

  • MMLU-PRO

    MMLU-PRO Raw · выше — лучше

    Результат

    0,4903

    Open LLM v2 · архив

    Публичная таблицадругая ревизия

  • BBH

    BBH Raw · выше — лучше

    Результат

    0,6391

    Open LLM v2 · архив

    Публичная таблицадругая ревизия

  • MUSR

    MUSR Raw · выше — лучше

    Результат

    0,4114

    Open LLM v2 · архив

    Публичная таблицадругая ревизия

  • IFEval

    IFEval Raw · выше — лучше

    Результат

    0,8064

    Open LLM v2 · архив

    Публичная таблицадругая ревизия

Что умеет модель

Описание составлено по карточке модели и метаданным Hugging Face

Qwen2.5-Lumen-14B — языковая модель на архитектуре Qwen2.5, созданная сообществом и основанная на Qwen/Qwen2.5-14B-Instruct. Модель дообучали методами предпочтений DPO и ORPO на примерах художественного письма и обратной связи, затем объединили несколько чекпойнтов и вариантов SLERP и Model Stock. Она предназначена для следования инструкциям, написания историй и ролевых диалогов.

Модель содержит 14 млрд параметров, работает с текстом и поддерживает контекст до 32 768 токенов. В карточке приведён англоязычный пример генерации; метаданные также указывают английский язык. Репозиторий содержит веса Safetensors, а для модели доступны сторонние варианты GGUF. Базовая лицензия — Apache-2.0.

Модель подходит для творческой генерации и диалогов, однако опубликованные сведения не содержат систематических оценок качества или гарантий стабильного поведения; приведённый пример зависит от выбранных параметров генерации.

Сценарии применения

  • написание художественных рассказов и глав романа
  • продолжение заданного литературного отрывка
  • ролевые диалоги с вымышленными персонажами
  • генерация ответов на инструкции в формате чата

Ограничения

В карточке нет систематических результатов оценки качества, поэтому уровень надёжности для конкретных сценариев не установлен. Пример генерации приведён с конкретными параметрами выборки, которые могут влиять на стиль и связность ответа.

Для каких профессий подходит3

Разделы каталога профессий, в задачах которых используется эта модель

Все профессии

Карточка модели

README от автора на Hugging Face, как он его опубликовал

Открыть на Hugging Face

Qwen2.5-Lumen-14B

  • Qwen direct preference optimization finetuned for ~3 epochs.

wCcJkdrVDUH6m0AN9Lv3B~2.png

A qwen2.5 preference finetune, targeting prompt adherence, storywriting and roleplay.


  • Llama.cpp

(GGUF) Thanks QuantFactory

(GGUF) Thanks mradermacher

(GGUF) Thanks Triangle104

Other quant repositories also exist on huggingface and can be searched for.

Training Notes

Trained Qwen2.5-14B-Instruct for 2 epochs on NVidia A100, and on dataset jondurbin/gutenberg-dpo-v0.1, saving different checkpoints along the way (completely different runs at varying epochs and learning rates).

Tanliboy trained Qwen2.5-14B-Instruct for 1 epoch on HuggingFaceH4/ultrafeedback_binarized, (Credit to Tanliboy! Check out the model here)

Mass checkpoint merged, Based on Qwen2.5-14B-Instruct (Base Model).

Merge

  • Merged with a sophosympatheia's SLERP gradient "Ultrafeedback-Binarized DPO" and "Gutenberg DPO"

  • Merged with a sophosympatheia's SLERP gradient "Qwen2.5-14B-Instruct" and "Gutenberg DPO"

  • Merged all DPO checkpoints and SLERP variations with MODEL_STOCK to analyze geometric properties and get the most performant aspects of all runs/merges. Model Stock was chosen due to the similarity between the merged models.

  • This was chosen due to the fact that evaluation for ORPO is unclear, so it's hard to know which runs are the best.

One-Attempt generated example:

  • Temp 1.3 [1], Min_P 0.012 [4], TFS 0.97 [3], Smooth_Factor 0.3 [2], Smoothing_Curve 1.1, Rep 1.1, Rep Range 1000

Title: A Harp for the Heart

Chapter One: Homebound

England, Winter 1893— As the steam-powered ferry made its way across the turbulent waters of the Channel, each tremor that reverberated through the vessel felt like a calloused reminder to Edgar Thornton. His mind was already conjuring visions of his beloved London—the sprawling city draped in an ever-present blanket of melancholic grey. Yet within him lingered more pressing memories: the whizzing sound of artillery shells as they streaked across Afghan deserts, leaving behind billowing black smoke and devastating wreckage.

Edgar's injuries were visible, the ghostly scar of a saber cut marring one cheek; less evident, however, was the toll that service took on his spirit. Each warbling note from an opera played by some amateur on the ferry sliced into Edgar's hollow heart, echoing through him with a bitter lament—the song was Adagio and it meant Slow down. Only now did he dare allow himself to slow—decades since music had flowed through the seams of his body in this way.

For years he had been sought out, from England to New York, as the country's foremost harpist, not just for his fingers nimble on a range of scales but also for how each piece conveyed an insurmountable melancholy—a reflection of his solitary life and unending loneliness. Each recital saw thousands of faces that remembered his smiles or silences; yet today he had no one to return them to.

The ferry's horn blared as they approached Dover, cutting through Edgar’s reverie much the same as cannonade did in Kabul. He thought about London with mixed emotions—longing for the grandeur and splendor of its concert halls, a city vibrant under lamplight but devoid of comfort or real life; a home where he was no less alien than anywhere else.

A soft murmur drew his gaze to a young woman standing near the window, her back to him. Her blonde hair cascaded in waves, shimmering under the lantern light as if lit from within. Despite the chill of the approaching winter night, she appeared oblivious—entirely entranced by something on the far horizon or lost in thought.

His breath caught at a sensation that had been uncommon for years but familiar to his boyhood—a light flutter tickling just behind his ribcage. And as her shoulders drew back and her chin tipped upwards, he saw a hint of the same unspoken story reflected in her eyes: dreams big enough to leave them scarred by distance and time.

And then she spoke without ever turning around—

"Do you hear that?" she asked into the gathering darkness outside their window. "Is that London?"[...]



The night was pitch black with only the thinnest sliver of a moon slicing across the sky, barely enough to paint the horizon a sickly shade of greyish-blue. Below it stretched an expanse of rolling hills and forests that swayed like shadows in an eternal nightmare. The wind carried with it not the scents of living earth but the taint of the bloodied battlegrounds below, where for too long had men, women, and beasts clashed and bled.

Ariath rode amidst this gloomy spectacle atop a pale stallion named Ebon, a beast as weary and war-weary as she was. Her dark cloak billowed around her like an inky cloud, contrasting starkly with the barren landscape through which they traveled. She had not chosen such darkness—nor any choice at all; tonight, every decision from waking to fighting lay in someone else's cold, indifferent hands.

Ebon’s hooves stirred the heavy mud of long trampled fields, sinking silently into the earth as he treaded towards her rallying army. It was said that this night they would fight, not for honor nor country—no, those noble phantoms had vanished many a moon past—but to stand once more between the realm and its approaching destruction.

Her chestnut mare, a beauty in years past before every bone seemed like it could break from its own weight, whinnied quietly under Ariath's calming voice. Her friends awaited not far ahead; they were an eclectic bunch—half-brother Raen, the steadfast knight who fought by her side for almost all their battles, Elaeril the mage known more as a wielder of shadow spells than life-giving magic, and finally Meryn the roguish, whose sword was as quick as his wit.[...]


  • Temp 1.3, Min_P 0.012, Rep 1.1

Chapter Seven

It was early morning, but the heavy clouds of melancholy still clung tightly to Anna's heart like a shroud over her spirit. The atmosphere outside mirrored this burden; rain fell softly, soaking the flower beds as mournful streaks rolled lazily across the gray sky in muted tones of despair.

Anna could not have taken her steps outwards to face the world until she had come to some kind of resolution within herself, even if it were no more than acceptance. She reflected deeply over how far away everything now seemed from what was once close and familiar. In every leaf trembling on a wet branch, in the low sighs of a wind that wandered lost without a purpose across the gray fields, she felt echoes of her inner turmoil.

She remembered her younger days. Life had then been filled with smiles, warmth and acceptance — her friends called her Annaarina, the spirit they saw in her so vividly, bright-eyed and kind. Their little gathering at Kitty’s had been an annual celebration for all four families: Oblonskys’, Shtcherbats’ys’, Karenins’, and Count Vronsky’s parents — each as familiar and loved a presence in that room as Anna.

But where were their faces now? Where was her happiness from those times, which only seemed to taunt the present?

"Here we are again,” she murmured in self-accusation with a mocking bitterness. The first snow had fallen softly upon them last winter then — and today it reminded her of something too lovely and cruel; something not allowed now in the rigid confines that life had chosen as its own path.

Determined to avoid being weighed down further, Anna prepared herself mentally to face Dolly, her dearest friend amongst women who seemed like a beacon among dim lights. In moments such as these she believed fully that somehow they could ease or even rid this dark fog from her heart — yet Anna felt apprehension nevertheless. Would their kinship and bond stand unchanged in light of recent circumstances? Perhaps now it would reveal cracks that Dolly was oblivious to, since the old Annaarina had been seen so rarely even then.


The carriage jolted through wet cobblestones as Anna finally made her way to Darya Alexandrovna’s house. The house stood there as warmly and welcomingly familiar with smoke still rising from chimneys, signaling signs of comfort within. Anna pulled off a shivering hand — dampness and cold seeped in merciless through fine silks.

Sneezing lightly into lace-lined white gloves, Anna steadied herself before rapping at the door. There would be no more room for her to hide; any attempt at self-preservation now lay broken between her heart and truth.

Anna stepped inside, greeted by the sight of Dolly busying about an unswept floor. For a fleeting moment, those familiarly busy hands paused — brows furrowed gently in concern even before their full attention met Anna.

"Why so early? Is there something...?" began Dolly warmly yet tentatively as she approached to take the delicate gloved hand.

Anna bit her lip at Darya's solicitous face. Would she now have to divulge her secrets of how utterly unhinged everything had become, or simply hint enough for suspicion?

"I — I need your advice," Anna whispered shyly, for there was no courage left to handle the world straight-up that day. Dolly hugged the fragile figure and replied earnestly, "Speak with me here freely. No matter what is troubling you..."

That simple reassurance gave Anna courage. The despairing weight lessened as though carried a little by this soft-hearted soul.

“I have changed more than you would ever understand...” began Anna softly, but the words broke as unshed tears finally welled. "Your eyes tell their story alone," soothed Dolly, patting gently her friend’s hand. “Do not be ashamed.”

In the wake of support and open hearts, Anna continued her revelation, pouring out all fears and hesitations — she who had been so content to live life sheltered inside, was suddenly exposed in full force by necessity alone; fragile glass cast into a brutal world.


At intervals throughout that day, Anna could scarcely believe how swiftly and faithfully Darya’s love proved the eternal rock to stand against crumbling shores. Amidst shared tea and comfort that night though shadows still lingered darkly between them.

When asked about Count Vronsky's presence with the family at gatherings or personal times now, Anna met this inquiry with a breathy sorrowful silence which spoke volumes to both the depth of their new estrangement, as well as how truly beloved the name was yet by Anna still in spite all waning prospects for resolution.

Daria continued to encourage her friend steadfastly; even shared letters and small consolations to penultimate messages destined to Count Vronsky alone. Each offered another grain's worth of hope for peace - both with him, herself, and beyond her personal dilemmas within that cloaked society so mercilessly turning its back against any form of irregularities in married life.

By the end of their meeting late evening shadows were upon them once more, and Anna reluctantly made to depart — a newly found strength (but still sorely tested) rippling across the troubled soul. However, as Anna prepared herself for returning to her solitary chambers, Dolly rose swiftly off the settee with concern laced in her tone.

“But we have barely scratched the surface of what might truly be affecting your heart.” The sympathy etched upon her face almost spoke a plea for understanding. Anna reached over and took hands of compassionate warmth — eyes wide with gratefulness.

"I think perhaps it is all right to stay here for the night… while I gather myself," she confided quietly into Darya’s caring ears.[...]


As you can see the model has mostly adapted to the intended response style from Gutenberg dataset.

Recipe

models:
  - model: v000000/Qwen2.5-14B-Gutenberg-1e-Delta
  - model: v000000/Qwen2.5-14B-Gutenberg-0.6e-Sequential
  - model: v000000/Qwen2.5-14B-Gutenberg-0.25e-Early
  - model: v000000/Qwen2.5-14B-Gutenberg-2e-Sequential
  - model: v000000/Qwen2.5-14B-Gutenberg-0.37e-Early
  - model: v000000/Qwen2.5-14B-Gutenberg-2e-Zeta
  - model: v000000/Qwen2.5-14B-Gutenberg-1e-Theta
  - model: tanliboy/lambda-qwen2.5-14b-dpo-test
  - model: v000000/Qwen2.5-14B-Gutenberg-1e-Delta
  - model: tanliboy/lambda-qwen2.5-14b-dpo-test
  - model: v000000/Qwen2.5-14B-Gutenberg-UltraLambda-Slerpeno
  - model: v000000/Qwen2.5-14B-Gutenberg-Instruct-Slerpeno
base_model: v000000/Qwen2.5-14B-Gutenberg-1e-Delta
merge_method: model_stock
dtype: bfloat16

If your use case is character-based roleplay, please consider using the prompts below for an enhanced experience

Finetune and merge

This is a merge and finetune of pre-trained language models.

Models Merged

Arxiv 2403.19522

The following models were included in the merge:

  • v000000/Qwen2.5-14B-Gutenberg-1e-Delta
  • v000000/Qwen2.5-14B-Gutenberg-0.6e-Sequential
  • v000000/Qwen2.5-14B-Gutenberg-0.25e-Early
  • v000000/Qwen2.5-14B-Gutenberg-2e-Sequential
  • v000000/Qwen2.5-14B-Gutenberg-0.37e-Early
  • v000000/Qwen2.5-14B-Gutenberg-2e-Zeta
  • v000000/Qwen2.5-14B-Gutenberg-1e-Theta
  • v000000/Qwen2.5-14B-Gutenberg-UltraLambda-Slerpeno
  • v000000/Qwen2.5-14B-Gutenberg-Instruct-Slerpeno
  • tanliboy/lambda-qwen2.5-14b-dpo-test

  • Context Length: Full 131,072 tokens and generation 8192 tokens
  • Qwen2(ChatML) Prompt format

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

MetricValue
Avg.32.20
IFEval (0-Shot)80.64
BBH (3-Shot)48.51
MATH Lvl 5 (4-Shot)0.00
GPQA (0-shot)10.40
MuSR (0-shot)10.29
MMLU-PRO (5-shot)43.36

Как запустить

На своей GPU в ailist. Цена — по текущим предложениям провайдеров

GPU в ailist

Универсальный рантайм · transformers

Можно запустить

Подбираем GPU…

Класс архитектуры
Qwen2ForCausalLM
Ревизия весов
a02b06b

0 моделей