hahahafofo's picture
Update README.md
3685230
|
raw
history blame contribute delete
1.71 kB
---
license: apache-2.0
---
模型使用在生成稳定扩散(Stable Diffusion)的提示语(prompt)。它主要通过中文生成相应的英文提示语,充分利用Qwen的能力,支持包括古诗词在内的多种形式进行提示语的生成。此模型基于2,000条数据进行特定于任务的微调(SFT),训练过程中,经历了5个训练周期(Epoch),采用了1e-05的学习率(Learning Rate),并最终达到了0.19的损失值(Loss)。
模型的对应代码已发布于GitHub,您可以通过以下链接访问并使用SDXL-Turbo进行快速测试生成的提示语:
[https://github.com/zhongpei/Qwen-SDXL-Turbo.git](https://github.com/zhongpei/Qwen-SDXL-Turbo.git)
This model is designed to generate prompts for Stable Diffusion. It primarily translates Chinese into corresponding English prompts, leveraging the capabilities of Qwen, and supports a variety of formats including classical poetry for prompt generation. The model has been fine-tuned specifically for this task (SFT) using a dataset of 2,000 entries. During its training phase, it underwent 5 epochs with a learning rate of 1e-05 and achieved a loss value of 0.19.
The corresponding code for the model is available on GitHub. You can access and use SDXL-Turbo for rapid test generation of prompts through the following link:
[https://github.com/zhongpei/Qwen-SDXL-Turbo.git](https://github.com/zhongpei/Qwen-SDXL-Turbo.git)
**参数表格(Parameters Table):**
| 参数 | 值 |
| -------------- | -------- |
| 数据量(Data) | 2,000条 |
| 训练周期(Epochs) | 5 |
| 学习率(Learning Rate) | 1e-05 |
| 损失值(Loss) | 0.19 |