loss=nan - Githubissues

locuslab / TCN

Sequence modeling benchmarks and temporal convolutional networks

https://github.com/locuslab/TCN

MIT License

4.13k stars 875 forks source link

Open uchiha99999 opened 4 months ago

uchiha99999 commented 4 months ago

Why did I lose=nan after training for a while?

My input and output channels are 30, and the sequence length is 10