likeatingcake

Results 3 comments of likeatingcake

Hello, I added some data to your weight training and trained according to step 1 and step 2. Finally, I got a weight, but when I used this weight for inference, I found that the generated image was very blurry. When I used your weight, it was normal again.

> 您好，我想请问一下您使用其他权重推理是改动了什么地方，我最近重新训练的模型权重在推理时都是产生很严重的噪音，除了官方给的模型以外其他效果都很差，不知道是不是我改动的问题。我是改了loop.py文件的权重路径，如何进行推理测试的 您好，请问问题解决了吗，我也是在loop.py里修改为我自己训练的模型权重的，但是inference出来的效果很差，但在train时生成的验证的图片是正确的啊

Hello, I added some data to your weight training and trained according to step 1 and step 2. Finally, I got a weight, but when I used this weight for inference, I found that the generated image was very blurry. When I used your weight, it was normal again.

> 我仅仅只是进行了技术测试，个人感觉如果要看到较好的效果的话，应该要10w以上的数据。你还需要在作者提供的预训练模型上接着训练。 > […](#) 请问可以分享一下新的inference代码吗？

Hello, I added some data to your weight training and trained according to step 1 and step 2. Finally, I got a weight, but when I used this weight for inference, I found that the generated image was very blurry. When I used your weight, it was normal again.

> > > 您好，我想请问一下您使用其他权重推理是改动了什么地方，我最近重新训练的模型权重在推理时都是产生很严重的噪音，除了官方给的模型以外其他效果都很差，不知道是不是我改动的问题。我是改了 loop.py 文件的权重路径，如何进行推理测试的 > > > > > > 您好，请问问题解决了吗，我也是在 loop.py 里修改为我自己训练的模型权重的，但是inference出来的效果很差，但在train时生成的验证的图片是正确的啊 > > > > 我当时训练时监控训练和验证，显示一切正常。但使用所得训练权重去测试时却出现了跟你这个差不多的情况，无论怎么更改inference代码也没有改变结果 好的谢谢你！！