Views
No views yet
| 类别 | 样本数量 | 说明 |
|---|---|---|
| Math（数学） | 36,568 | 数学推理和计算问题 |
| Exam（考试） | 2,432 | 各类考试题目 |
| STEM（理工科） | 12,648 | 科学、技术、工程、数学领域 |
| General（通用） | 58,352 | 弱智吧、逻辑推理、小红书、知乎、Chat等多元场景 |
| 总计 | 110,000 | - |
1from transformers import AutoModelForCausalLM, AutoTokenizer
2
3# 加载模型和分词器
4model_name = "suyu-io/Llama-3.1-8B-Thinking-Distill-R1"
5tokenizer = AutoTokenizer.from_pretrained(model_name)
6model = AutoModelForCausalLM.from_pretrained(
7 model_name,
8 torch_dtype="auto",
9 device_map="auto"
10)
11
12# 推理示例
13prompt = "请解释一下量子纠缠的原理"
14messages = [
15 {"role": "user", "content": prompt}
16]
17
18inputs = tokenizer.apply_chat_template(
19 messages,
20 return_tensors="pt"
21).to(model.device)
22
23outputs = model.generate(
24 inputs,
25 max_new_tokens=2048,
26 temperature=0.7,
27 top_p=0.9,
28 do_sample=True
29)
30
31response = tokenizer.decode(outputs[0], skip_special_tokens=True)
32print(response)1from unsloth import FastLanguageModel
2
3model, tokenizer = FastLanguageModel.from_pretrained(
4 model_name = "suyu-io/Llama-3.1-8B-Thinking-Distill-R1",
5 max_seq_length = 2048,
6 dtype = None,
7 load_in_4bit = True,
8)
9
10FastLanguageModel.for_inference(model)
11
12inputs = tokenizer(
13 "请详细解释一下相对论",
14 return_tensors="pt"
15).to("cuda")
16
17outputs = model.generate(
18 **inputs, # 解包字典
19 max_new_tokens = 2048,
20 use_cache = True,
21 temperature = 0.2,
22 min_p = 0.2,
23 repetition_penalty = 1.1
24)
25
26print(tokenizer.decode(outputs[0], skip_special_tokens=True))
271@misc{llama-3.1-8b-thinking-distill-r1,
2 author = {suyu-io},
3 title = {Llama-3.1-8B-Thinking-Distill-R1},
4 year = {2026},
5 publisher = {HuggingFace},
6 howpublished = {\url{https://huggingface.co/suyu-io/Llama-3.1-8B-Thinking-Distill-R1}}
7}