Skip to main content
stacey
Projects
deep-drive
Reports
Training Loss is worse when we use resnet24

Training Loss is worse when we use resnet24

Scott Condron
Created on December 16|Last edited on December 16
Here's a little analysis of training loss beside my other configurations.

Section 1


0.0000.0010.0020.0030.0040.0050.0060.0070.0080.0090.0100.0110.012learning...alexnetresnet152resnet18resnet34resnet50encoder2345678910111213141516batch_sizenull0.00.51.01.52.02.53.03.54.0train_loss
1
0.51.01.52.02.53.03.5train_loss0.30.40.50.60.70.8traffic_acc
Run set
14

Hmm... maybe we should try a higher learning rate?!
💡

e=mc2e=mc^2
  • Try lower learning rate
Scott Condron
Scott Condron •  
traffic_acc v. train_loss
Cool chart!
Reply
Scott Condron
Scott Condron •  
should try a higher learning rate Good idea! Maybe we can run a Hyperparameter search?
Reply

Created with ❤️ on Weights & Biases.

https://wandb.ai/stacey/deep-drive/reports/Training-Loss-is-worse-when-we-use-resnet24--VmlldzoxMzUwMjE5
Made with Weights & Biases. Sign up or log in to create reports like this one.