Skip to content

Merge with dev#26

Merged
hsson merged 65 commits intomasterfrom
dev
Apr 20, 2017
Merged

Merge with dev#26
hsson merged 65 commits intomasterfrom
dev

Conversation

@hsson
Copy link
Copy Markdown
Contributor

@hsson hsson commented Apr 20, 2017

No description provided.

Jesper Jaxing and others added 30 commits April 3, 2017 13:28
merge
db3ec23
added method to add secondary output layer
bc9154d
implemented pre-training on subreddits
1afbf9f
create methods for handling pre-training data
95b779e
non-working commit to allow for debugg help.
4afbf98
changed which variable was used to check if pre-training
193811d
only add sec_output when pre-train and add an extra layer after
e5be638
changed config to not include unnecissary parameters
d94d187
cleaned main.py by adding it to builder
fa9859c
sec_output uses softmax since a title only have one subreddit
1de1133
refactored add output method to one method
5c2db6c
@hsson @Jaxing
Adds the option to choose between GRU and LSTM units
2374ed4
@hsson @Jaxing
Adds a missing variable name change
f46c8ce
@hsson @Jaxing
Changes default dataset in config template
634d79b
merge
d73a5c4
added method to add secondary output layer
468d07d
implemented pre-training on subreddits
2e957a3
create methods for handling pre-training data
796a81d
non-working commit to allow for debugg help.
81c149d
changed which variable was used to check if pre-training
f5c49ed
only add sec_output when pre-train and add an extra layer after
8c5cb78
cleaned main.py by adding it to builder
7383a0b
sec_output uses softmax since a title only have one subreddit
157987d
refactored add output method to one method
90086ab
Merge branch 'feature/pre-train-subreddit' of github.com:kandidat-hig…
2561556 
…hlights/model into feature/pre-train-subreddit
@hsson
Refactors away redundant function in model builder
65cf80c
@hsson
Removes un-used constant
fbe203f
@hsson
Removes unused epoch counting for pre-training
ac40573
@hsson
Merge pull request #15 from kandidat-highlights/feature/pre-train-sub…
8158619 
…reddit

Merges feature/pre train subreddit
@hsson
Removes redundant printing of cross entropy error
5f2d82a 
This removes the printing of cross entropy for training and validation error to the terminal. This is since the erros can still be visualised using tensorboard. It also makes the code more effective as it doesn't have to compute error twice.
Jesper Jaxing and others added 29 commits April 6, 2017 17:53
chande to actual variable
fc8a0e2
@Mxiim
Merge pull request #21 from kandidat-highlights/refactoring/subreddit…
73b5a23 
…-input

Refactoring/subreddit input
@Mxiim @Jaxing
Fixes bug when we would log twice for pretraining and actual training
ee1a015
@Mxiim @Jaxing
Fixes bug when we couldnt use concat-input AND pretraining for a model
3358a0e
@Jaxing
only make one call to label_vector
515489e
@jonatanalmen
Adds docker support
b856b66 
adds readme on how to run docker.
@hsson
Automatically run all configs if none is specified
8319891
@Jaxing
still need to fill in all ranges
c1726bb
@Jaxing
all but the datasets
fcd3bb8
@Jaxing
added dataset
112a5d1
@Mxiim @Jaxing
Removes wrong files and adds two more learning rates
f600402
@Jaxing
fixed bugg
a8f23e6
@hsson @Jaxing
Changes user count to take UNK user into account
97997d2
@hsson @Jaxing
Removes dataset that doesn't exist
762e8f0
@hsson @Jaxing
Increases interval between hidden neuron sizes
d01a683
@hsson @Jaxing
Increases prediction limit intervals
c2bc3ef
@hsson @Jaxing
Tries more common batch sizes
4bed82e
@hsson @Jaxing
USe consistent step between lstm neurons
8883e78
@hsson @Jaxing
Uses dropout probabilities from original dropout paper
16acae6
@hsson @Jaxing
Uses new hyperparameters from dev
51bde47
@hsson @Jaxing
Adds more possible l2 factors
72fdb3b
@Mxiim @Jaxing
Adds 0 for hidden layer, adds 400 for lstm_neurons and removes 100 fr…
9e68557 
…om lstm_neurons
@Mxiim @Jaxing
Fixes indentation
50c14aa
@Jaxing
network name is guaranteed unique
45d6370
@hsson
Adds a script that automatically downloads the data needed
d479982
@hsson
Refactors all double precision floating points to single precision to…
97f0ffb 
… save memory
@Mxiim
Merge pull request #25 from kandidat-highlights/refactor/floating-pre…
2e9a3b4 
…cision

Memory improvements
@Mxiim
Resets graph even if it failed
233b090
Changed the header to include pre_train_subreddit
0a35092
@hsson hsson merged commit 44a2db0 into master Apr 20, 2017
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@hsson @Mxiim @jonatanalmen