jcjohnson/torch-rnn issues and pull requests

#75 - Explanation of hyperparameters?

Issue - State: open - Opened by atomicthumbs over 8 years ago - 7 comments

#74 - (Optionally) Stream sampled text

Pull Request - State: open - Opened by ChrisCummins over 8 years ago

#73 - (optionally) Save and restore the checkpoint iteration counter

Pull Request - State: closed - Opened by ChrisCummins over 8 years ago - 1 comment

#72 - Epoch and iteration state are lost when loading from checkpoint

Issue - State: closed - Opened by ChrisCummins over 8 years ago - 2 comments

#71 - Sampling: incremental printing of generated text

Issue - State: open - Opened by gwern over 8 years ago - 2 comments

#70 - About Multi-threading

Issue - State: closed - Opened by already-taken-m17 over 8 years ago - 2 comments

#69 - cannot find lua package "optim"

Issue - State: open - Opened by ghazel over 8 years ago - 1 comment

#68 - Comparing 'likeliness' of some input against trained model

Issue - State: open - Opened by ChrisCummins over 8 years ago - 5 comments

#67 - Invalid multinomial distribution (sum of probabilities <= 0)

Issue - State: closed - Opened by AlekzNet over 8 years ago - 1 comment

#66 - Bi-Directional RNN support

Pull Request - State: closed - Opened by SeanNaren over 8 years ago - 12 comments

#65 - Gradually increasing calculation time

Issue - State: closed - Opened by AlekzNet over 8 years ago - 3 comments

#64 - Modified preprocess.py to accept syllabic prediction...

Pull Request - State: open - Opened by dreavjr over 8 years ago - 6 comments

#63 - The initial bias of the forget gate

Issue - State: closed - Opened by AlekzNet over 8 years ago - 2 comments

#62 - Batchnorm vs dropout

Issue - State: open - Opened by AlekzNet over 8 years ago - 3 comments

#61 - HTTP server for generating samples on demand

Pull Request - State: open - Opened by Hrant-Khachatrian over 8 years ago - 1 comment

#60 - Element-Research rnn

Issue - State: open - Opened by nicholas-leonard over 8 years ago - 8 comments

#59 - Checkpoints not being written

Issue - State: open - Opened by rubinovitz over 8 years ago - 10 comments

#58 - error on `train.lua`: expected near ')' at line 579

Issue - State: open - Opened by nylki over 8 years ago - 33 comments

#57 - Avoiding the index being out of bound

Pull Request - State: open - Opened by zhen-he over 8 years ago

#56 - cannot convert 'struct THCudaTensor *' to 'float'

Issue - State: closed - Opened by alexlenail over 8 years ago - 2 comments

#55 - How to use a trained char-based RNN language model?

Issue - State: closed - Opened by coodoo over 8 years ago - 1 comment

#54 - Bi-Directional RNN support

Pull Request - State: closed - Opened by SeanNaren over 8 years ago - 1 comment

#53 - [feature request] Character-level features

Issue - State: open - Opened by casey over 8 years ago

#52 - support UTF-8 start_text

Pull Request - State: open - Opened by maraoz over 8 years ago - 4 comments

#51 - Add support for bidirectional RNNs

Issue - State: open - Opened by iamalbert over 8 years ago - 1 comment

#50 - Memory allocation/leak issue

Issue - State: closed - Opened by buckohfive over 8 years ago - 5 comments

#49 - Implementing seq2seq learning?

Issue - State: closed - Opened by c0nn3r over 8 years ago - 1 comment

#48 - Install tutorial for Mac users

Issue - State: closed - Opened by jeffThompson over 8 years ago - 2 comments

#47 - Using non-ASCII characters in start_text crashes sample.lua

Issue - State: closed - Opened by ostrosablin over 8 years ago - 16 comments

#46 - Is it possible to save checkpoints without resetting/clearing the state of the model?

Issue - State: closed - Opened by AlekzNet over 8 years ago - 4 comments

#45 - Improved bookkeeping of settings when resuming from existing checkpoint.

Pull Request - State: closed - Opened by billzorn over 8 years ago

#44 - Sample h5 and json files don't work

Issue - State: closed - Opened by jeffThompson over 8 years ago - 2 comments

#43 - scripts/preprocess.py should sort tokens lexicographically

Issue - State: closed - Opened by AlekzNet over 8 years ago - 1 comment
Labels: duplicate

#42 - Modified the initial size of self.buffer3 in LSTM.lua

Pull Request - State: closed - Opened by zhen-he over 8 years ago - 1 comment

#41 - Modified the initial size of self.buffer2 in VanillaRNN.lua

Pull Request - State: closed - Opened by zhen-he over 8 years ago - 3 comments

#40 - init_from fails

Issue - State: closed - Opened by stephenpardy over 8 years ago - 2 comments

#39 - libhdf5-dev unavailable on latest Ubuntu?

Issue - State: closed - Opened by Patashu over 8 years ago - 3 comments

#38 - Stateful RNN mod in sampling

Issue - State: closed - Opened by aliabbasjp over 8 years ago - 2 comments
Labels: question

#37 - Updated training?

Issue - State: open - Opened by drzax over 8 years ago - 8 comments
Labels: enhancement

#36 - Fixed LSTM to return self.gradInput

Pull Request - State: closed - Opened by iassael over 8 years ago - 1 comment

#35 - Can torch-rnn be used with 0 test and validation data?

Issue - State: closed - Opened by AlekzNet over 8 years ago - 4 comments

#34 - torch-rnn does not make use of multiple CPU cores

Issue - State: closed - Opened by AlekzNet over 8 years ago - 13 comments

#33 - Resume training from checkpoints

Issue - State: closed - Opened by gwern over 8 years ago - 5 comments

#32 - Cutorch illegal memory access was encountered

Issue - State: closed - Opened by nomeyer over 8 years ago - 1 comment

#31 - dropout enabled during sampling!

Issue - State: closed - Opened by mrmartin over 8 years ago - 8 comments

#30 - Update preprocess.py to support Python 3

Pull Request - State: closed - Opened by mattbierner over 8 years ago - 2 comments

#29 - preprocess.py does't handle unicode character sequences correctly

Issue - State: closed - Opened by mattbierner over 8 years ago - 3 comments

#28 - sample.lua flag of -sample 0 raises error

Issue - State: closed - Opened by buckohfive over 8 years ago - 2 comments

#27 - preprocess.py uses np.uint32 when there are 256 items in the vocabulary

Issue - State: closed - Opened by casey over 8 years ago - 2 comments

#26 - scripts/preprocess.py--encoding flag not documented

Issue - State: closed - Opened by casey over 8 years ago - 2 comments

#25 - attempt to call method 'resetSize' (a nil value)

Issue - State: closed - Opened by rossgoodwin over 8 years ago - 6 comments

#24 - Sampling ouputs garbage, encoding problem?

Issue - State: closed - Opened by wohnjayne over 8 years ago - 4 comments

#23 - Training stops at 10th iteration

Issue - State: closed - Opened by rprechelt over 8 years ago

#22 - Added missing -num_layers description in flags.md and corrected a typo

Pull Request - State: closed - Opened by gevangelopoulos over 8 years ago - 1 comment

#21 - How to extend this code to deal with variable length squences?

Issue - State: closed - Opened by jundengdeng over 8 years ago - 3 comments

#20 - GRU cells

Pull Request - State: open - Opened by guillitte over 8 years ago - 8 comments

#19 - Can I add softmax layer at the top?

Issue - State: closed - Opened by didw over 8 years ago - 3 comments

#18 - Thank you !

Issue - State: closed - Opened by fusiller over 8 years ago - 5 comments

#17 - Are the defaults supposed to be a 1:1 match for char-rnn?

Issue - State: closed - Opened by clausd over 8 years ago - 4 comments

#16 - Fix `torch-hdf5` clone URL

Pull Request - State: closed - Opened by urish over 8 years ago - 1 comment

#15 - lua 5.2 compatibility fix

Pull Request - State: closed - Opened by soumith over 8 years ago - 2 comments

#14 - Fix model_type param in README example

Pull Request - State: closed - Opened by ageitgey over 8 years ago - 1 comment

#13 - I have a question in util.gradcheck

Issue - State: open - Opened by spk921 over 8 years ago - 1 comment
Labels: bug

#12 - Option for non HDF5

Pull Request - State: open - Opened by northanapon over 8 years ago - 1 comment

#11 - OpenCL backend slower than CPU

Issue - State: open - Opened by mewo2 over 8 years ago - 13 comments
Labels: bug

#10 - Sampling can't be run in CPU mode without cunn and cutorch

Issue - State: closed - Opened by youwayx over 8 years ago

#9 - Remove unnecessary cunn and cutorch dependencies

Pull Request - State: closed - Opened by youwayx over 8 years ago

#8 - Stream sampled text

Pull Request - State: open - Opened by maraoz over 8 years ago - 3 comments

#7 - Correct preprocessing for UTF-8 input files

Pull Request - State: closed - Opened by maraoz over 8 years ago

#6 - Use dirname instead of basename

Pull Request - State: closed - Opened by youwayx over 8 years ago - 1 comment

#5 - Checkpoint saving error when writing to json file

Issue - State: closed - Opened by youwayx over 8 years ago

#4 - Maybe need a `paths.mkdir('cv')` somewhere?

Issue - State: closed - Opened by hughperkins over 8 years ago - 1 comment

#3 - Instructions step 2, should be only one hyphen before each option I think?

Issue - State: closed - Opened by hughperkins over 8 years ago - 1 comment

#2 - Nuance for setup: `sudo apt-get install -y python2.7-dev`

Issue - State: closed - Opened by hughperkins over 8 years ago - 3 comments

#1 - For info, I think can install torch-hdf5 as `hdf5`?

Issue - State: closed - Opened by hughperkins over 8 years ago - 4 comments

GitHub / jcjohnson/torch-rnn issues and pull requests