richardyang92 / burn-examples

some neural networks implemented by burn framework

burn-examples

some neural networks implemented by burn framework:

1. bigram

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.

2. fashion mnist

runing cmd: cargo run --package fashion_mnist

training result: the model works well.

3. yolo v1

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.