Model Card for wangzihaogithub/job-educational-parser

This model is a fine-tuned version of job-educational-parser. It has been trained using TRL.

模型任务:输入岗位描述,输出岗位中要求的学历(如 "博士、硕士、本科"),遵循从高到低

训练数据:https://huggingface.co/datasets/wangzihaogithub/job-educational-parser-dataset-08-0-0805.

例如: 输入

{
  "model": "wangzihaogithub/job-educational-parser",
  "messages": [
    {
      "role": "system",
      "content": "从岗位中提取学历"
    },
    {
      "role": "user",
      "content": "游戏美术实习-原画类(26届提供转正)游戏类型&风格:欧美卡通 休闲类游戏 工作职责: 1. 配合各项目组美术需求(包括不限于原画,UI和动画类需求)的落地和整合; 2. 具有比较扎实的手绘能力,能够独立完成运营活动所需的美术设计需求; 3. 通过对主流AI产品的学习,总结提示词使用技巧,通过具体案例验证方法的有效性,协助团队建立规范化的AI应用方法论和完善AI工作流程。 岗位要求: 1.面向游戏/动画/数媒/雕塑/美术/工业设计等设计相关专业; 2.本科及以上学历; 3.2026年应届毕业生; 4.有优秀的绘画基础,会用手绘板,熟练掌握PS等2D设计软件 加分项: 1.美术院校相关专业者优先; 2.爱玩游戏者优先; 3.有相关美术设计实习或者工作经验者优先考虑; 4.性格乐观爽朗,善于表达。"
    }
  ],
  "max_tokens": 32
}

输出(固定格式):博士、硕士、本科

响应速度:100~200毫秒之间(显卡RTX3060,精度BF16)

准确率:98%, 评测集:https://huggingface.co/datasets/wangzihaogithub/job-educational-parser-dataset-08-0-0805/viewer/annotated-test.

Requirements

transformers>=4.51.0

Framework versions

  • TRL: 0.19.0
  • Transformers: 4.53.0
  • Pytorch: 2.6.0
  • Datasets: 3.6.0
  • Tokenizers: 0.21.2

Quick start

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_name = "wangzihaogithub/job-educational-parser"

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

print('model load success')

messages = [
    {"role": "system", "content": '从岗位中提取学历'},
    {"role": "user", "content": "游戏美术实习-原画类(26届提供转正)游戏类型&风格:欧美卡通 休闲类游戏 工作职责: 1. 配合各项目组美术需求(包括不限于原画,UI和动画类需求)的落地和整合; 2. 具有比较扎实的手绘能力,能够独立完成运营活动所需的美术设计需求; 3. 通过对主流AI产品的学习,总结提示词使用技巧,通过具体案例验证方法的有效性,协助团队建立规范化的AI应用方法论和完善AI工作流程。 岗位要求: 1.面向游戏/动画/数媒/雕塑/美术/工业设计等设计相关专业; 2.本科及以上学历; 3.2026年应届毕业生; 4.有优秀的绘画基础,会用手绘板,熟练掌握PS等2D设计软件 加分项: 1.美术院校相关专业者优先; 2.爱玩游戏者优先; 3.有相关美术设计实习或者工作经验者优先考虑; 4.性格乐观爽朗,善于表达。"}
]

text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
    enable_thinking=False
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

# 生成推理结果(关闭梯度计算)
with torch.inference_mode(), torch.amp.autocast('cuda'):
    generated_ids = model.generate(
        **model_inputs,
        max_new_tokens=32,# 输出结果为学历,不需要太多
        eos_token_id=tokenizer.eos_token_id,
        pad_token_id=tokenizer.pad_token_id,
        use_cache=True,
        do_sample=False,  # 是否采样(False 为 greedy decode)
        temperature=None, # 关闭
        top_p=None,# 关闭
        top_k=None# 关闭
    )
output_ids = generated_ids[0][len(model_inputs.input_ids[0]):].tolist()

# parsing thinking content
try:
    # rindex finding 151668 (</think>)
    index = len(output_ids) - output_ids[::-1].index(151668)
except ValueError:
    index = 0

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

print(content)

Training procedure

This model was trained with SFT.

Citations

Cite TRL as:

@misc{vonwerra2022trl,
    title        = {{TRL: Transformer Reinforcement Learning}},
    author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
    year         = 2020,
    journal      = {GitHub repository},
    publisher    = {GitHub},
    howpublished = {\url{https://github.com/huggingface/trl}}
}
Downloads last month
10
Safetensors
Model size
0.6B params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for wangzihaogithub/job-educational-parser

Finetuned
(490)
this model

Dataset used to train wangzihaogithub/job-educational-parser