In sentiment analysis of the IMDb reviews, which of the following techniques are used to speed up the training process? Choose all that applies. Group of answer choices

A. We can mask the padding tokens so that the training process will ignore them.
B. For all the rest (~43000) uncommon words, we can put them in 1000 OOV buckets.
C. Truncate each review to keep only the first few hundred characters since usually the first few sentences summarize the sentiment
D. Only keep the first 10000 most common words