Open xnghu opened 3 years ago
When I tested with my hardware, PyTorch 1.8.1 with CUDA 11.1 wass faster without jit than using 1.7.1 with.
In my experience, the jit model was 10-20% faster. Feel free to test it, if it is really faster anyways with torch 1.8.1 without jit then we can just remove that warning and change the defaults.
I'm on a 3090 and so can't use CUDA 10 for tests, but my guess is that improvements in CUDA 11.1 outweigh any benefits from jit + CUDA 11.0
Is it possible to somehow generate a jit model for a newer version?
我已经能够用这个代码产生一些惊人的结果,非常感谢这个项目,这太棒了。
当我运行代码时,我得到的消息是:
Setting jit to False because torch version is not 1.7.1
我目前一直在使用Python 3.8.6和torch 1.8.1 + cu102。
我还没有完全理解jit是什么,但它听起来像是提高了代码的速度?我应该降级到火炬1.7.1吗?
How to solve this problem
I've been able to generate some astounding results with this code, thank you SO much for this project, it's fantastic.
when I run the code i get the message:
Setting jit to False because torch version is not 1.7.1
I currently have been using Python 3.8.6 with torch 1.8.1+cu102.
I have yet to fulling understand what jit is, but it sounds like it make improve the speed of the code? should I downgrade to torch 1.7.1?