vita-group / gan-slimming Goto Github PK
View Code? Open in Web Editor NEW[ECCV 2020] "All-in-One GAN Compression by Unified Optimization" by Haotao Wang, Shupeng Gui, Haichuan Yang, Ji Liu, and Zhangyang Wang
License: MIT License
[ECCV 2020] "All-in-One GAN Compression by Unified Optimization" by Haotao Wang, Shupeng Gui, Haichuan Yang, Ji Liu, and Zhangyang Wang
License: MIT License
We use the model in (https://drive.google.com/drive/folders/1aBAyuaGX1jxF3lqx1HEzEieX_cXjIapq,horse2zebra, GS32, GS8), but we can't get the results as the results in your paper. The results for GS32 as following:
The results for GS8 as following:
I want to calculate the fid of these models of your paper. Can you give some suggestions?
Hello, thanks for sharing the code.
I trained the cycleGAN model on horse2zebra dataset for several times without the quant. But the best FID is about 97, I wonder if I should change some hyperparameters when training the model without quant?
And could you please upload the horse2zebra model as a reference, thank you very much!
看了论文和gs.py, 给我的感觉是是基于cyclegan生成的图像, 外加剪枝loss重新训练一个pix2pix的G. 我这样的理解有问题吗?
既然是衡量student_output_img和teacher_output_img的差距, 为什么感知loss是强于mse的呢?
我的理解是训练目标就单纯是让G_p2p(student)的结果拟合G_cyclegan(teacher)的结果, mse这样直接的约束为何没有感知loss这样更加偏向语义的约束效果好呢?
Thank you best work,i meet some confusion about dense model.why need dense model?
The dataset is already one-to-one.
I think we can skip directly to the fourth step without a dense model,
but I don’t know if my idea is correct. Hope you can help me.
Thanks for your geart work! I have traind my cycle gan model with the official CycleGAN codes, but when I run gs.py the error showed :
RuntimeError: Error(s) in loading state_dict for Generator:
Missing key(s) in state_dict: "model.2.weight", "model.2.bias", "model.5.weight", "model.5.bias", "model.10.conv_block.2.weight", "model.10.conv_block.2.bias", "model.11.conv_block.2.weight", "model.11.conv_block.2.bias", "model.12.conv_block.2.weight", "model.12.conv_block.2.bias", "model.13.conv_block.2.weight", "model.13.conv_block.2.bias", "model.14.conv_block.2.weight", "model.14.conv_block.2.bias", "model.15.conv_block.2.weight", "model.15.conv_block.2.bias", "model.16.conv_block.2.weight", "model.16.conv_block.2.bias", "model.17.conv_block.2.weight", "model.17.conv_block.2.bias", "model.18.conv_block.2.weight", "model.18.conv_block.2.bias", "model.20.weight", "model.20.bias", "model.23.weight", "model.23.bias".
I didn't change the official cycle gan codes, and the options I trained my model is "python train.py --dataroot ./datasets/Task12_BBR2color --name BBR2color --model cycle_gan --pool_size 50 --no_dropout --gpu_ids 1 --preprocess scale_width_and_crop --load_size 1920 --crop_size 360 --batch_size 1 --display_port 2020"
The pytorch version is 1.7
Hi,
Could you please explain steps for how to Generate style transfer results on the training set? Which file do we need to run?
Do I need to again train your pre-trained models? I used the GS-32 pre-trained dense model. I am not getting good results. Could you please help me?
With Regards,
Jatin Kumar
What an exciting work! I noticed two details when running the code.
(1) Whether it is GS or fine-tune, G and D use a smaller learning rate, while the norm factor gamma uses a larger learning rate. Why?
(2) When running gs.py, only dataset horse2zebra ( in 104 lines) needs to load the pre training model and retrain. Why?
Hi, thanks for sharing the code.
Can you make available the compression code for the CartoonGan model?
Hi! @htwang14
Thanks for your nice work! I have seen the Conv2dQuant and ConvTrans2dQuant operators in the in models/models.py file, but the feature map before the quantized convolution and weights are still floating-point values.
How can I get an int8 quantified model, it can be saved with 8-bit.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.