C3W4 Assignment Exercise 4 Samese model for question duplicates: Evaluation part

Course Q&A Natural Language Processing NLP with Sequence Models
1

Hi, I have trouble on C3W43Assignment Exercise 4 Samese model for question duplicates: Evaluation part exercise 4:
There the default test data created by default:
test_gen = tf.data.Dataset.from_tensor_slices(((test_Q1, test_Q2),None)).batch(batch_size=batch_size)

However, when I perform evaluation with already test data by utilize test_gen, I am stuck
Here is my part of pseuduo code:

for i in range(0, len(test_Q1)):
element = next(iter(test_gen))

How can I use element with a batched data to feed in the model ?
I think:
v1, v2 = model(q1, q2)

However how to get q1, q2 correct as input to the model ?

Zhiyi

When creating a post, please add:

  • Week # must be added in the tags option of the post.
  • Link to the classroom item you are referring to:
  • Description (include relevant info but please do not post solution code or your entire notebook)
2

I noticed output of next(iter(test_gen)) is tuple include test_Q1, and test_Q2 information already, but still not get correct in model input parameters.

3

Are you posting in the right specialization?

4

Hi @Zhiyi_Li2

You do not need the for loop. The test_gen is an iterable object and you can pass it to model.predict(..).

And as you later realized, the test_gen is already a tuple, so you don’t have to worry about wrapping (test_Q1, test_Q2) as it is already in the test_gen.

Cheers

5

Thanks [arvyzukai]
I think loop still needed, test_gen is an iterable object,
still need get next item as:

instead of model.predict(), for prediction, should I use model(test_gen.get_next()) instead?

6

test_gen has 2 elements one for (test_Q1, test_Q2), another for None, only test_gen.get_next()[0] has (test_Q1, test_Q2).

7

Errors for model prediction part for model((x1,x2))
x, y = iterator.get_next()
x1, x2 = x
# print ("x1: ", x1)
# print ("x2: ", x2)
v1, v2 = model((x1, x2))

Error message is: too many values to unpack (expected 2)

When I run model to predict for test data, input should be test_Q1, test_Q2. How can I get input correct ?

8

Problem for C3W3_Assignment Duplicate questions exercise 4 errors:

I need create v1, v2 based on the Siamese model: Here is my snippet of code:

for elem in test_gen:
v1, v2 = model.predict(elem[0])

The error complains as: too many values to unpack (expected 2)

I think model.predict(…) function expect a tuple (Question 1, Question 2). However, elem[0] is exactly from test_gen, can help to fix this too many values to unpack errors ?

9

I would say test_gen has nested tuple ((q1, q2), None).

10

Hi @Zhiyi_Li2

Yes, that is what I meant, by that you don’t need to pack Q1 and Q2 again. In the train_dataset it was ((q1, q2), 1), but here instead of 1, we use None.

Did you solve your issue? What was the problem if you did?

11

Problem is how to specify predict=model.predict(…) input parameters ? Currently I use input test_gen like:

v1, v2 = model.predict(test_gen)
The system complains with:

too many values to unpack (expected 2)

What exact I put input parameters here ?

12

I used the code:
v1, v2 = model.predict(test_gen, verbose=0)

Still same errors:
ValueError: too many values to unpack (expected 2)

For first input parameter, should I need do some change? from 512 dimension(batch size) to 2 instead? How ?

13

I use the code:
pred = model.predict(next(iter(test_gen)), verbose=0)

The system complains cannot handle None type as errors:
Inputs to a layer should be tensors. Got ‘None’ (of type <class ‘NoneType’>) as input for layer ‘SiameseModel’.

I think next(iter(test_gen)) has a size 2, one for (q,1,q2), another for None, It cause the system cannot handle None type, What is really input for pred = model.predict() is? second parameter verbose = 0 already done, what is first parameter ?