some neural networks implemented by burn framework:
runing cmd:
if you want to train this model, please run cargo run --package bigram train
, or you can just run cargo run --package bigram generate
to generate some characters.
training result: accuracy is about 35% for 10 epochs training, quite normal for this simplest language model.
runing cmd: cargo run --package fashion_mnist
training result: the model works well.
runing cmd: cargo run --package yolo_v1
training result: Training process can't be completed as the loss stops decreasing at around 14, leading to its termination by the burn framework. I'm still investigating the reasons behind this.