bentrevett
2b666b3cba
update to torchtext 0.9
2021-03-12 13:25:47 +00:00
bentrevett
ba82b9231e
updated tutorials + readme with latest versions of libs.
2021-02-17 12:52:57 +00:00
bentrevett
492b6d572a
added batch_first to remove a permute
2019-11-06 14:01:23 +00:00
bentrevett
2753e93ca4
reran all notebooks with latest pytorch and torchtext to ensure still working
2019-09-16 11:32:18 +01:00
bentrevett
808aaafa6f
removed extra print statement added whilst debugging. changed default model back to 2d cnn
2019-04-22 14:27:48 +01:00
bentrevett
807ec2082a
fixed typos in max pool figure and size of tensors after convolutional filters ( #35 )
2019-04-22 13:26:13 +01:00
bentrevett
be122f3a6f
added model.eval() in predict sentiment functions ( #31 )
2019-04-10 10:27:20 +01:00
bentrevett
669a9b8436
removed erroneous statement about the cnn model being the fastest model
2019-04-01 16:40:55 +01:00
bentrevett
525ae0b416
lots of formatting changes
2019-03-29 16:57:00 +00:00
bentrevett
eea44cb487
added missing unk init and fixed some formatting to notebook 4
2019-03-29 16:37:21 +00:00
bentrevett
59920c56cf
removed incorrect xavier init example from notebook 2
2019-03-23 14:45:52 +00:00
bentrevett
bf8cc46e48
added parameter count and epoch timer functions to all notebooks. also added cnn1d imlementation to notebook 4
2019-03-21 22:48:21 +00:00
bentrevett
cd09d20b65
fixed out of glove vector initialization and missing generate bigrams functions in fasttext implementation
2019-03-21 17:22:25 +00:00
bentrevett
6e037d8386
fixed wording and updated assets for notebook 4
2019-03-10 16:28:21 +00:00
bentrevett
61bf7da1d0
Update for TorchText 0.3
...
- reworded a lot of the explanations
- fixed a lot of typos
- updated some images and moved new ones to assets folder
- fixed deterministic evaluation with `torch.backends.cudnn.deterministic = True`
- tensor_type in Fields becomes dtype
- changed train, valid, test TorchText dataset objects to be named train_data, valid_data and test_data
- TorchText iterators now have `repeat = False` by default, so removed
- TorchText iterators now take a `torch.device` instead of an int representing the GPU number (and -1 for CPU)
- `F.sigmoid` is now `torch.sigmoid`
- padding sentences in the CNN model is now fixed (before spaCy would tokenize the <pad> token into three separate tokens)
2018-10-17 17:54:31 +01:00
bentrevett
9f71d9d5e8
fixed typo in CNN model (thanks Remy) and changed size of tensor comments to use n_filters instead of n_channels, reflecting the actual variable names used
2018-07-19 17:23:16 +01:00
bentrevett
8fc3a75721
typo in conv tutorial about calculating output size of conv layers
2018-07-05 17:16:10 +01:00
bentrevett
f34d972d74
added conv tutorial
2018-07-05 17:12:00 +01:00
