magicboomliu / accelerator-simple-template Goto Github PK
View Code? Open in Web Editor NEWThis is a simple template using HuggingFace Accelerator for DDP-training/Saving/Loading/Pushing.
License: MIT License
This is a simple template using HuggingFace Accelerator for DDP-training/Saving/Loading/Pushing.
License: MIT License
感谢作者的复现,我尝试在KITTI数据集上微调,直接运行的作者复现的training代码,不同的地方是我修改了下dataset,但是总是显示 torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 20.00 MiB (GPU 0; 23.67 GiB total capacity; 23.15 GiB already allocated; 10.25 MiB free; 23.27 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF,而且无论怎么调低分辨率都会显示out of memory,请问作者知道是什么原因吗?
另外作者方便加一下微信不,我的微信号:shaoshuweifighting, 十分感谢。
left_image_data_resized in the training code is normalized to [0, 1] and fed to the vae encoder. However, according to Marigold, the vae encoder accepts data range [-1, -1]. Is this on purpose?
Thanks for sharing the training scripts! I have two questions about the training data. In the data files, I noticed the use of occlusion files. Do they belong to the sub-dataset called Disparity Occlusion Weights? Also, for training on KITTI, it appears that only RGB images are used without depth supervision. I'm a bit confused about this.
Hi, first, thanks much for sharing the training codes.
I found a few differences between your reproduction and the original paper. In the original paper, multi-resolution noise is adapted for a significant performance improvement. However, your coding uses torc.randn_like() to produce noise.
Is my understanding correct or not? Actually, I am new to the diffusion model. If yes, is there any next plan to implement the multi-resolution noise in your codes?
I'm looking forward to hearing back from you.
Best regards.
I plan to add the controlnet into the original Marigold, once it have been tested, I will upgrade the code.
Thank for your training code!Can you share how to implement the hybrid training in the original paper? Whether to map them to the same depth space
Hello,
Thank you for making the training code available. I am interested in seeing how the model performs on scene flow datasets. Would it be possible for you to share any test results that you have conducted on the datasets?
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.