库名称：transformers
许可证：apache-2.0
数据集：

• attn-signs/kolmogorov-3
• attn-signs/russian-code
  语言：
• ru
  基础模型：
• Qwen/Qwen3-8B

Qwen3-8B-ru

• [英文]
  基于Qwen3的模型，适配俄语文本生成任务。
• [俄文]
  Qwen3的微调版本，专为生成俄语文本优化。

模型详情 / 模型细节

• [英文]
  在2块NVIDIA A100 GPU上进行了12小时的LoRA监督微调，训练1个epoch，使用数据集：
  attn-signs/kolmogorov-3；
  attn-signs/russian-code；

• [俄文]
  LoRA SFT训练在两块NVIDIA A100上完成，耗时约12小时。
  完整遍历数据集：
  attn-signs/kolmogorov-3；
  attn-signs/russian-code；

模型描述

• 开发者：[Reisen Raumberg (Attention Signs团队)]
• 支持语言（NLP）：[俄语/英语]
• 微调基础模型：[Qwen3]

采用DeepSpeed（第三阶段）、HF.Accelerator进行分布式训练和融合AdamW优化器。
GPU耗时：NVIDIA A100 12小时

训练使用HuggingFace Accelerator结合Microsoft DeepSpeed（第三阶段）分配参数和优化器状态，以及融合AdamW。
GPU小时数：NVIDIA A100 12小时

训练配置

[模型]  
模型名称或路径 = "Qwen/Qwen3-8B"  

[数据集]  
数据集 = [  
    'attn-signs/kolmogorov-3',  
    'attn-signs/russian-code',  
]  
数据集比例 = [  
    1,  
    1  
]  
测试集比例 = 0.05  
对话字段 = "conversation"  
生成评估样本 = false  
评估策略 = "steps"  
评估步长 = 500  
数据加载器工作线程数 = 2  
移除未使用列 = true  

[运行]  
保存策略 = "steps"  
保存步长 = 500  
保存总数限制 = 3  
运行名称 = "sft-qwen3-8b"  
报告至 = "wandb"  
记录初始步骤 = true  
记录步长 = 1  
输出目录 = "models/attn-signs-qwen3-8b"  
项目名称 = "sft-qwen3"  

[训练]  
仅训练完成部分 = true  
每设备训练批次大小 = 1  
每设备评估批次大小 = 1  
训练周期数 = 1  
学习率 = 0.00004  
梯度累积步数 = 8  
梯度检查点 = true  
预热步数 = 10  
bf16 = true  
随机种子 = 42  
使用peft = true  
最大长度 = 4096  

[融合]  
使用liger = true  
注意力实现 = "flash_attention_2"  

[lora]  
lora目标模块 = [  
    "k_proj",  
    "v_proj",  
    "q_proj",  
    "o_proj",  
    "gate_proj",  
    "up_proj",  
    "down_proj",  
]  
lora秩 = 512  
loraα值 = 512  

[分词器]  
助手消息模板 = "<|im_start|>assistant"  
填充标记 = "<|endoftext|>"  
结束标记 = "<|im_end|>"  
对话模板 = "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0].role == 'system' %}\n        {{- messages[0].content + '\\n\\n' }}\n    {%- endif %}\n    {{- \"# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0].role == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0].content + '<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}\n{%- for message in messages[::-1] %}\n    {%- set index = (messages|length - 1) - loop.index0 %}\n    {%- if ns.multi_step_tool and message.role == \"user\" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}\n        {%- set ns.multi_step_tool = false %}\n        {%- set ns.last_query_index = index %}\n    {%- endif %}\n{%- endfor %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {%- set content = message.content %}\n        {%- set reasoning_content = '' %}\n        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}\n            {%- set reasoning_content = message.reasoning_content %}\n        {%- else %}\n            {%- if '</think>' in message.content %}\n                {%- set content = message.content.split('</think>')[-1].lstrip('\\n') %}\n                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\\n').split('<think>')[-1].lstrip('\\n') %}\n            {%- endif %}\n        {%- endif %}\n        {%- if loop.index0 > ns.last_query_index %}\n            {%- if loop.last or (not loop.last and reasoning_content) %}\n                {{- '<|im_start|>' + message.role + '\\n<think>\\n' + reasoning_content.strip('\\n') + '\\n</think>\\n\\n' + content.lstrip('\\n') }}\n            {%- else %}\n                {{- '<|im_start|>' + message.role + '\\n' + content }}\n            {%- endif %}\n        {%- else %}\n            {{- '<|im_start|>' + message.role + '\\n' + content }}\n        {%- endif %}\n        {%- if message.tool_calls %}\n            {%- for tool_call in message.tool_calls %}\n                {%- if (loop.first and content) or (not loop.first) %}\n                    {{- '\\n' }}\n                {%- endif %}\n                {%- if tool_call.function %}\n                    {%- set tool_call = tool_call.function %}\n                {%- endif %}\n                {{- '<tool_call>\\n{\"name\": \"' }}\n                {{- tool_call.name }}\n                {{- '\", \"arguments\": ' }}\n                {%- if tool_call.arguments is string %}\n                    {{- tool_call.arguments }}\n                {%- else %}\n                    {{- tool_call.arguments | tojson }}\n                {%- endif %}\n                {{- '}\\n</tool_call>' }}\n            {%- endfor %}\n        {%- endif %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if loop.first or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- message.content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n    {%- if enable_thinking is defined and enable_thinking is false %}\n        {{- '<think>\\n\\n</think>\\n\\n' }}\n    {%- endif %}\n{%- endif %}"  

使用示例

from transformers import AutoModelForCausalLM, AutoTokenizer  

model_name = "attn-signs/Qwen3-8B-ru"  

# 加载分词器和模型  
tokenizer = AutoTokenizer.from_pretrained(model_name)  
model = AutoModelForCausalLM.from_pretrained(  
    model_name,  
    torch_dtype="auto",  
    device_map="auto"  
)  

# 准备模型输入  
prompt = "给我简要介绍一下大语言模型。"  
messages = [  
    {"role": "user", "content": prompt}  
]  
text = tokenizer.apply_chat_template(  
    messages,  
    tokenize=False,  
    add_generation_prompt=True,  
    enable_thinking=True # 切换思考模式与非思考模式。默认为True。  
)  
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)  

# 执行文本生成  
generated_ids = model.generate(  
    **model_inputs,  
    max_new_tokens=32768  
)  
output_ids = generated_ids[0][len(model_inputs.input_ids[0]):].tolist()   

# 解析思考内容  
try:  
    # 反向查找151668 (</think>)  
    index = len(output_ids) - output_ids[::-1].index(151668)  
except ValueError:  
    index = 0  

thinking_content = tokenizer.decode(output_ids[:index], skip_special_tokens=True).strip("\n")  
content = tokenizer.decode(output_ids[index:], skip_special_tokens=True).strip("\n")  

print("思考内容:", thinking_content)  
print("生成内容:", content)