Open sekkit opened 4 weeks ago
In principle, it is no different from the current method; it simply increases the context window, which consequently requires more computational resources.
In principle, it is no different from the current method; it simply increases the context window, which consequently requires more computational resources.
But I found no train scripts in your repo, so I have to write it myself? or u mean just change the inference's context window to achieve that? Sorry I'm new to AI, learning it now.
The training code is not released now. It will be released once paper acceptance.
How to train high faces' model, any idea on the dataset's format and resource requirement?