Hello everyone,
While I was helping a learner, I noticed one my submission model had different percentage before and after submission. So just wanted to know people who have cleared their assignment, while in the assignment it must shown the model passed with some percentage and after submission of model.h5 showed a different percentage.
This is to know how grader assessment can differ from assignment notebook we do.
So for me, Before I submitted the assignment my passing percentage was 75% and after submission it was 84%.
I wanted to know if people have got higher to this percentage. Even lower one who have passed the grader can put their input. So that we know what makes the model more better as we can compare each others models. But do not share any codes.
Thank you for being part of this discussion.
Regards
DP
Hi DP,
It might be the case that different data is tested on the assignment and the grader (even the case of different images, if its a vision problem), so the accuracy will depend on the data being fed to test it!
Hi Gent,
Thank you for replying. I know the images might be different. But these assignment uses a premed data and I think all the learners are using the same data but can’t guarantee if the assignment was updated.
I only wanted to see if the model algorithm varied too much for this assignments.
Regards
DP
1 Like
Hello fellows can anybody help me as I am getting this error each and every time after I submit my model.h5 file.
“There was a problem grading your submission. Check stderr for more details.”
Hello Ejaz,
did you try a different browser or different computer once you faced this issue. Make sure you have deleted your cache and browser history when you retry doing the assignment and submission.
In case if you are still facing the issue let me know.
Regards
DP
1 Like
Hello there Deepti firstly I was submitting my assignment on Google Chrome but I also tried submitting it through Microsoft Edge but still getting the same error
“There was a problem grading your submission. Check stderr for more details.”
along with Grade Score of 0 whereas in code my Grade score is 79.99363365798419.
Hello Ejaz,
Kindly click my name and then message to send your notebook. kindly send a download ipnyb notebook rather than sharing a link to your collar link.
Regards
DP
1 Like
Hello Ejaz,
-
In defining basic convolution block, data format, need to used with the recalled function and not global variable name. Also for the 2nd convolution block you need write data format again for that layer too.
-
In the pad the input image to have dimensions to the nearest power of two, the input shape is incorrect. Use the input shape as per you have used in upsampling step for original size in the later cell correctly in the fcn8 decoder cell.
-
In the FCN8, block 1, 2, 3 need to have kernel size, pooling_size and pool_stride of 2. Mention it as below
x = conv_block(x, 32, 2, 2, 2)
-
Remove use_bias from the all the Upsample steps.
-
Again in the below code line, data_format need to recalled with define function in the cell.
load the pool 3 prediction and do a 1x1 convolution to reshape it to the same shape of o above
-
For the compile model step, incorrect loss used. Also for learning rate, kindly use learning_rate=1E-2 this way and run the training model
-
While selecting number epoch in Exercise 5, check the expected output mentions what number epoch needs to be used. Kindly use the same number of epoch.
Once your issue is resolved, kindly let me know. Sorry for the delay in response.
Regards
DP
1 Like
Thank you so much for the response and your time. I will let you know after my issue is resolved.
Hello Deepti, did you get my code which I sent you?
I already responded to you
1 Like