[GGUF/Q6_K] Finetune версия Qwen3, адаптированная для генерации русского текста.

tools

ollama run OxW/Qwen3-8b-ru-i1

curl http://localhost:11434/api/chat \
  -d '{
    "model": "OxW/Qwen3-8b-ru-i1",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='OxW/Qwen3-8b-ru-i1',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'OxW/Qwen3-8b-ru-i1',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Applications

Claude Code ollama launch claude --model OxW/Qwen3-8b-ru-i1

Codex ollama launch codex --model OxW/Qwen3-8b-ru-i1

OpenCode ollama launch opencode --model OxW/Qwen3-8b-ru-i1

OpenClaw ollama launch openclaw --model OxW/Qwen3-8b-ru-i1

Models

View all →

Name

1 model

Size

Context

Input

Qwen3-8b-ru-i1:latest

6.7GB · 40K context window · Text · 10 months ago

Qwen3-8b-ru-i1:latest

6.7GB

40K

Text

Readme

Source: HF

Qwen3-8B-ru

[EN]
Qwen3-based model, adapted for russian text generation tasks.
[RU]
Finetune версия Qwen3, адаптированная для генерации русского текста.

Model Details / Детализация модели

[EN]
LoRA supervised finetuning was performed on 2xA100 NVIDIA GPUs for 12h for 1 epoch on datasets:
attn-signs/kolmogorov-3;
attn-signs/russian-code;
[RU]
LoRA SFT цикл был выполнен на двух NVIDIA A100, обучение длилось около 12 часов.
Прогон полной эпохи датасетов:
attn-signs/kolmogorov-3;
attn-signs/russian-code;