license, language, tasks, base_model, base_model_relation, tags

license	language	tasks	base_model	base_model_relation	tags
Apache License 2.0	zh	text-generation	Qwen/Qwen2.5-7B-Instruct	finetune	lora llama-factory

qwen2.5-75b-lora

该模型是在法律案件数据集上对 ./Qwen/Qwen2.5-7B-Instruct 进行微调后的版本。

• Loss： 0.4874
• eval_loss：0.48537

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

Training hyperparameters

train

• per_device_train_batch_size: 8
• gradient_accumulation_steps: 4
• learning_rate: 2.0e-4
• num_train_epochs: 2.0
• lr_scheduler_type: cosine
• warmup_ratio: 0.1
• bf16: true
• ddp_timeout: 180000000
• resume_from_checkpoint: null

eval

• eval_dataset: my_dataset_test
• per_device_eval_batch_size: 4
• eval_strategy: steps
• eval_steps: 500 # 调小评估步数，确保训练中执行多次评估
• metric_for_best_model: eval_loss # 指定核心评估指标
• load_best_model_at_end: true # 训练结束后加载最优模型

Training results

Training Loss	Epoch	Step	Validation Loss
0.8000	0.45	500	0.5190
0.4995	0.90	1000	0.4995
0.4874	1.36	1500	0.4910
0.4121	1.81	2000	0.4854
0.4874	2.00	2212	0.4854

---

当前模型的贡献者未提供更加详细的模型介绍。模型文件和权重，可浏览“模型文件”页面获取。

您可以通过如下git clone命令，或者ModelScope SDK来下载模型

SDK下载

#安装ModelScope
pip install modelscope

#SDK模型下载
from modelscope import snapshot_download
model_dir = snapshot_download('tongzang/Qwen2.5-7b-lora-law')

Git下载

#Git模型下载
git clone https://www.modelscope.cn/tongzang/Qwen2.5-7b-lora-law.git

如果您是本模型的贡献者，我们邀请您根据模型贡献文档，及时完善模型卡片内容。