---
title: RankMe_FinInd
canonical_url: "https://www.modelscope.cn/models/BAAI_Industry_Competition_FinEcon_dev/RankMe_FinInd"
md_url: "https://www.modelscope.cn/models/BAAI_Industry_Competition_FinEcon_dev/RankMe_FinInd.md"
repository: BAAI_Industry_Competition_FinEcon_dev/RankMe_FinInd
last_updated: 2024-12-30
license: "Apache License 2.0"
base_model:
  - Qwen/Qwen2.5-7B-Instruct
base_model_relation: adapter
parameters: 40.4M
tensor_type:
  - F32
library_name:
  - lora
  - safetensors
downloads: 419
stars: 0
tags:
  - llama-factory
  - lora
---

# RankMe_FinInd

> RankMe_FinInd - BAAI_Industry_Competition_FinEcon_dev 在 ModelScope 开源的模型。RankMe_FinInd

BAAI_Industry_Competition_FinEcon_dev/RankMe_FinInd 是 ModelScope 魔搭社区上的 40.4M 参数机器学习模型，采用 Apache License 2.0 许可，基于 Qwen/Qwen2.5-7B-Instruct 构建。

- **Repository**: BAAI_Industry_Competition_FinEcon_dev/RankMe_FinInd
- **License**: Apache License 2.0
- **Parameters**: 40.4M
- **Base model**: Qwen/Qwen2.5-7B-Instruct
- **Tags**: llama-factory, lora
- **Downloads**: 419
- **Stars**: 0
- **Last updated**: 2024-12-30

Source: https://www.modelscope.cn/models/BAAI_Industry_Competition_FinEcon_dev/RankMe_FinInd

---

# RankMe_FinInd

基于[Qwen2.5-7B-Instruct](https://modelscope.cn/models/Qwen/Qwen2.5-7B-Instruct)微调的金融行业大模型

训练数据使用[IndustryInstruction_Finance-Economics_RankMe_Filter_Regen · 数据集](https://modelscope.cn/datasets/BAAI_Industry_Competition_FinEcon_dev/IndustryInstruction_Finance-Economics_RankMe_Filter_Regen/)

## 文件目录结构

```
.
├── adapter_config.json // 模型文件
├── adapter_model.safetensors // 模型文件
├── inference.py // 推理代码
├── RankMe_FinInd_12081411.jsonl // 推理结果
├── RankMe_FinInd_report.pdf // 报告
├── README.md
├── run_log.jsonl // 推理日志
└── train_config.yaml // 训练配置
└── trainer_log.jsonl // 训练日志
```

## 训练过程

### 超参数

- learning_rate: 1e-05
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 8
- total_train_batch_size: 32
- total_eval_batch_size: 4
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.05
- num_epochs: 2

### 结果

| Training Loss | Epoch  | Step | Validation Loss |
|:-------------:|:------:|:----:|:---------------:|
| 0.8448        | 0.1021 | 36   | 0.8503          |
| 0.7601        | 0.2043 | 72   | 0.8113          |
| 0.7435        | 0.3064 | 108  | 0.7970          |
| 0.7562        | 0.4085 | 144  | 0.7891          |
| 0.7248        | 0.5106 | 180  | 0.7842          |
| 0.7566        | 0.6128 | 216  | 0.7815          |
| 0.7442        | 0.7149 | 252  | 0.7784          |
| 0.7513        | 0.8170 | 288  | 0.7759          |
| 0.7469        | 0.9191 | 324  | 0.7747          |
| 0.7998        | 1.0213 | 360  | 0.7730          |
| 0.7227        | 1.1234 | 396  | 0.7717          |
| 0.7466        | 1.2255 | 432  | 0.7709          |
| 0.7677        | 1.3277 | 468  | 0.7701          |
| 0.7146        | 1.4298 | 504  | 0.7696          |
| 0.7407        | 1.5319 | 540  | 0.7691          |
| 0.7524        | 1.6340 | 576  | 0.7689          |
| 0.7149        | 1.7362 | 612  | 0.7687          |
| 0.7456        | 1.8383 | 648  | 0.7687          |
| 0.7433        | 1.9404 | 684  | 0.7688          |


### Framework versions

- PEFT 0.12.0
- Transformers 4.46.1
- Pytorch 2.3.1+cu121
- Datasets 3.1.0
- Tokenizers 0.20.2