search

bdhingra / tweet2vec

Twitter hashtag prediction
BSD 2-Clause "Simplified" License
281 stars 90 forks source link

mask_input #4

Closed meshiguge closed 8 years ago

meshiguge commented 8 years ago

hi, the mask_input was used in GRU layer l_mask = lasagne.layers.InputLayer(shape=(N_BATCH,MAX_LENGTH), input_var=mask, name='mask') according to this tutorial should N_BATCH be replaced with None? as the n_batchs are not equal in all batch ?

bdhingra commented 8 years ago

Yes, that can be replaced with None as the last batch in an epoch may not be equal to N_BATCH. I have not seen any issues running the code as it is though.

meshiguge commented 8 years ago

dear bdhingra, how about the MAX_LENGTH, every input sequence is also different, should I replace it also?

bdhingra commented 8 years ago

No, the second dimension of the input will always have size MAX_LENGTH. See this.