关于“TensorFlow Dataset API”的评价

Vikram M. · 评论over 2 years之前

I had some difficulty following along with this lab since the previous lab exercise was not included in this course. I would recommend adding the first lab in that folder to the course since future labs reference it.

Charles B. · 评论over 2 years之前

Olha B. · 评论over 2 years之前

Need more clarity on where to run the commands - on Terminal or just click on the arrow of the instructions window?

ep m. · 评论over 2 years之前

Alejandro A. · 评论over 2 years之前

Adán T. · 评论over 2 years之前

Santosh S. · 评论over 2 years之前

Rafael G. · 评论over 2 years之前

Stepan G. · 评论over 2 years之前

Pablo Leonardo L. · 评论over 2 years之前

Tomas B. · 评论over 2 years之前

Sonali R. · 评论over 2 years之前

Николай К. · 评论over 2 years之前

Gaurav P. · 评论over 2 years之前

Tejas T. · 评论over 2 years之前

Siwatchara S. · 评论over 2 years之前

Sharad Kumar G. · 评论over 2 years之前

quite messy unclear how to do tutorials with this part, and functions are not working as expected. lab task #2 --------------------------------------------------------------------------- AssertionError Traceback (most recent call last) Cell In[66], line 26 23 loss =loss_mse(X_batch, Y_batch, w0, w1) # TODO -- Your code here. 24 print(MSG.format(step=step, loss=loss, w0=w0.numpy(), w1=w1.numpy())) ---> 26 assert loss < 0.0001 27 assert abs(w0 - 2) < 0.001 28 assert abs(w1 - 10) < 0.001 AssertionError: part 4b AttributeError Traceback (most recent call last) Cell In[108], line 5 1 BATCH_SIZE = 2 3 tempds = create_dataset('../toy_data/taxi-train*', batch_size=2) ----> 5 for X_batch, Y_batch in tempds.take(2): 6 pprint({k: v.numpy() for k, v in X_batch.items()}) 7 print(Y_batch.numpy(), "\n") AttributeError: 'tuple' object has no attribute 'take' part 4c --------------------------------------------------------------------------- TypeError Traceback (most recent call last) Cell In[112], line 1 ----> 1 tempds = create_dataset('../toy_data/taxi-train*', 2, 'train') 2 print(list(tempds.take(1))) Cell In[111], line 6, in create_dataset(pattern, batch_size, mode) 2 def create_dataset(pattern, batch_size=1, mode='eval'): 3 dataset = tf.data.experimental.make_csv_dataset( 4 pattern, batch_size, CSV_COLUMNS, DEFAULTS) ----> 6 dataset = tf.data(pattern) # TODO -- Your code here. 8 if mode == 'train': 9 dataset = dataset.shuffle() # TODO -- Your code here. TypeError: 'module' object is not callable

Mika K. · 评论over 2 years之前

Walter D. · 评论over 2 years之前

Alfredo B. · 评论over 2 years之前

Marc N. · 评论over 2 years之前

W T. · 评论over 2 years之前

Anagha B. · 评论over 2 years之前

Sometimes during coding you just do not know what you have to do if you are not familiar with Tensorflow syntax. But still small enough steps to test and try!

Daniel S. · 评论over 2 years之前

Good practice, some instructions could've been more clear ( e.g. about desired way of implementing things like filtering feature columns, buffer size for shuffling ). Also it instructed to set "batch_size, column_names and column_defaults" in the first create_dataset which apparently wasn't actually desired ( batch_size shouldn't be set yet, only in the second time, otherwise there'll be an assertion error as it's expecting just scalars in the aserts )

Jasper v. · 评论over 2 years之前