Rickymomo/YangtzeHandBook

Rickymomo/ YangtzeHandBook

18 Downloads Updated 9 months ago

Based on the fine-tuning model of the student handbook of Yangtze University, the base model is Qwen3_8B

tools thinking 8b

ollama run Rickymomo/YangtzeHandBook:8b

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

from ollama import chat

response = chat(
    model='Rickymomo/YangtzeHandBook:8b',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

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

Applications

Claude Code ollama launch claude --model Rickymomo/YangtzeHandBook:8b

OpenCode ollama launch opencode --model Rickymomo/YangtzeHandBook:8b

Hermes Agent ollama launch hermes --model Rickymomo/YangtzeHandBook:8b

OpenClaw ollama launch openclaw --model Rickymomo/YangtzeHandBook:8b

Models

View all →

Name

4 models

Size / Usage

Context

Input

YangtzeHandBook:8b

16GB · 40K context window · Text · 9 months ago

YangtzeHandBook:8b

16GB

40K

Text

Readme

简介

基于Yangtze学生手册2023版，使用Easy_dataset+DeepSeek-reasoner生成两万条问答数据，经过AI质量评估与人工筛选，选出其中的一万条质量较高的单轮问答数据集进行全量微调训练。

8B版本的重要训练参数如下

训练资源：4*H100
微调方法：FULL
训练方式：SFT
学习率：0.00003
epochs：5
batch_size：40
gradient_accumulation_steps：8
截断长度：3000
Deepseed：stage3

运行8B全量版本

ollama run Rickymomo/YangtzeHandBook:8b

运行32B微调版本

ollama run Rickymomo/YangtzeHandBook:32B_Q4_K_M