Hi,
I have an issue my assignment runtime takes ages to just complete one epoch it took 856 seconds? any idea what is wrong?
Hi @Sajjad_Ali ,
It’s worth noting that, while it’s common for image datasets to take longer to process initially due to their size, this should be less of an issue here since the model uses a flattened input rather than convolutional layers.
Just make sure that your model architecture matches the instructed structure. Specifically, check that it has around 4,062,468 trainable parameters (and 4,096,580 in total) to confirm that it’s not over-parameterized.
Also, make sure your runtime is set to GPU and not CPU or TPU.
Hope it helps! Feel free to ask if you need further assistance.
I just ran my code, and each epoch takes approximately 1 minute.
1 Like
Hi @Alireza_Saei ,
Thanks for reply yeah all my parameters matches with the required parameters, and also when i run the notebook it gives me warning to run without GPU but when i check the runtime it is T4 GPU. I don’t know where is the it should take 28 to 30 s per epoch but 10 mnts per epoch there must be a problem.
Hi,
It’s worth double-checking other parts of the code to ensure they match the required setup. Sometimes minor differences in data preprocessing or batch sizes can also affect runtime. Additionally, try running your notebook in another Google account’s Colab environment. There’s a possibility that you’ve used up your current account’s available GPU quota and might need to wait for it to reset, which can limit GPU access speed.
Let me know if this helps!
1 Like
It’s best to troubleshoot on your own first, as it can help you better understand the potential issues. However, if you’re feeling stuck, feel free to share your notebook with mentors in a private message, and we can go through it together to identify any potential problems.
1 Like
check your steps per epoch codes for exercise 6, as in each epoch steps per epoch act as batches.
Also please make sure your mobilnet code is mentioned correctly.
there are older threads where you can cross check your codes.
1 Like
thanks @Deepti_Prasad and @Alireza_Saei
as for as my understanding I am unable to identify the bug or issue i will share my notebook in private.