Hello there
I have solved this problem with a slope rate if validation loss 0.00013 with a timespan averaging a minute per epoch. I figured it out with the following code snippets to build the model architecture using the dataset Sentiment140 dataset,. I want know what is your best slope rate and how to get or how to achieve the lowest loss figure rate and using what model architecture.
[code removed - moderator]
thank you in advance 
Model architecture is specific to a problem. You’ll have to try different architectures to figure out what works best for the dataset or use a pretrained model via transfer learning. See options like automl to perform neural architecture search on a dataset.
I tried everything, but still cannot get required slope. Please help!
This is the architechture I use:
[code removed - moderator]
PLEASE HELP!
same here, where you able to crack it. could you pls share some idea on good architecture for this assignment.
Please read about linregress and then look at techniques to ensure that training and validation losses are not too different in their trajectories. Remember that this assignment wants you to avoid overfitting. There’s no need to shoot for very high training accuracy.
Here’s the passing criteria: To pass this assignment the slope of your val_loss curve should be 0.0005 at maximum.