Output (GPU P100)

• step 0: train loss 4.5788, val loss 4.5842
• step 500: train loss 1.7982, val loss 1.9718
• step 1000: train loss 0.9245, val loss 1.5703
• step 1500: train loss 0.3320, val loss 1.8531
• step 2000: train loss 0.1062, val loss 2.3758
• step 2500: train loss 0.0717, val loss 2.6864
• step 3000: train loss 0.0624, val loss 2.9426
• step 3500: train loss 0.0577, val loss 3.0741
• step 4000: train loss 0.0545, val loss 3.2397
• step 4500: train loss 0.0530, val loss 3.2999

Generated Text

Interesaings 70 In the programme that the Railways person of the customers, will be organized at schemes etchess will be enhanced. Therefore, this year we have made a policy for reducing nine per cent this are all times likely to being fieline per cent and freement contries, retraining empties and growth rate we in in in introduced tickets in buing equipped with mpriove further, the dispaying new thichat the operfore dividend payment of Rs.274 crore to the Railways. Con

📚 References

• Attention Is All You Need (Transformer), NeurIPS 2017
• Deep Residual Learning for Image Recognition (ResNet), CVPR 2016