Commits

History for

Commits on Nov 16, 2018

keywords
Julian Kates-Harbeck
committed
8c48c7b

Commits on Apr 6, 2018

added support for multiple times to disruption ROC evaluation for shallow models

Julian Kates-Harbeck
authored and
Julian Kates-Harbeck
committed
a083b94

Commits on Mar 24, 2018

added functionality to compute multiple ROC values for different T_min values at once. Add option to use ProcessGenerator or not
Julian Kates-Harbeck
committed
9322202

Commits on Mar 22, 2018

added support for torch fully convolutional model. Temporal convolutions and spatial convolutions. TODO is to add multi GPU support and customizability via conf. Parameters currently hardcoded.
Julian Kates-Harbeck
committed
ad5de2f
Merge branch 'jdev' of https://github.com/PPPLDeepLearning/plasma-python into jdev

Julian Kates-Harbeck
authored and
Julian Kates-Harbeck
committed
e97f251
adding support for pytorch model

Julian Kates-Harbeck
authored and
Julian Kates-Harbeck
committed
8bf886a

Commits on Mar 21, 2018

printing roc values to log
Julian Kates-Harbeck
committed
6d25d37
whitespace
Julian Kates-Harbeck
committed
68eb287
implemented more general hash that allows deep copies of the conf to maintain the hash
Julian Kates-Harbeck
committed
8aa5c1a

Commits on Mar 19, 2018

supporting logging of val and test roc values for various T_min_warn times
Julian Kates-Harbeck
committed
1d5cf00

Commits on Mar 7, 2018

remove bidirectional RNN since it violates causality. Add resilience for when batch_normalization is not defined in the conf
Julian Kates-Harbeck
committed
76a566e
remove bidirectional RNN since it violates causality. Add resilience for when batch_normalization is not defined in the conf
Julian Kates-Harbeck
committed
735e1c2

Commits on Feb 9, 2018

modified printout to add walltime
Julian Kates-Harbeck
committed
e21f431
removed print
Julian Kates-Harbeck
committed
e06bcb1
implemented randomized cutting of first batch_size shots in the buffer. This means shots end at random times even at the beginning of training, which removes the bias that for the first couple of t…
Julian Kates-Harbeck
committed
b080734
not saving from within preprocessor class

Julian Kates-Harbeck
authored and
Julian Kates-Harbeck
committed
374e069
fixed bug where all workers had the same numpy random seed (it gets reset to 0 in preprocessor). Thus need to explicitly give each worker a different random seed right before training.
Julian Kates-Harbeck
committed
8ce98dc

Commits on Jan 12, 2018

Option to apply batch norm between (most) linear and non-linear layers, bidirectional wrapper for lstm
ASvyatkovskiy
committed
7d6fcb1

Commits on Dec 26, 2017

Switch to hashlib.md5 hashing
ASvyatkovskiy
committed
db55cbc

Commits on Dec 25, 2017

Fix TypeError in layer definitions in Python3
ASvyatkovskiy
committed
1e25e7c

Commits on Dec 15, 2017

slight model architecture modification. dense layer after convolutions has dense_size
Julian Kates-Harbeck
committed
67b1189

Commits on Dec 10, 2017

sorting of 1D signals such that they are at the end. Extended is_strictly_positive to 1D signals. Fixed bug where attribute and function had the same name is_strictly_positive
Julian Kates-Harbeck
committed
2e7b3fa

Commits on Nov 21, 2017

add option to predict from custom path for shallow and deep models
Julian Kates-Harbeck
committed
8533eec

Commits on Nov 18, 2017

max depth as 0 for XGBoost but None for decision trees

Julian Kates-Harbeck
authored and
Julian Kates-Harbeck
committed
b5a1902
creating model directory
Julian Kates-Harbeck
committed
3f70621
Merge branch 'master' of https://github.com/PPPLDeepLearning/plasma-python
Julian Kates-Harbeck
committed
83af19c
syntax
Julian Kates-Harbeck
committed
b7b653b
bug in model saving location

Julian Kates-Harbeck
authored and
Julian Kates-Harbeck
committed
4ff40e6

Commits on Nov 17, 2017

flexible T_warning for shallow models

Julian Kates-Harbeck
authored and
Julian Kates-Harbeck
committed
7a55743

Commits on Nov 16, 2017

adding multi layer perceptron to shallow models
Julian Kates-Harbeck
committed
35b5c48
added mlp as shallow model

Julian Kates-Harbeck
authored and
Julian Kates-Harbeck
committed
5049f57

Commits on Nov 15, 2017

Merge branch 'master' of https://github.com/PPPLDeepLearning/plasma-python
Julian Kates-Harbeck
committed
d047b98
syntax
Julian Kates-Harbeck
committed
c899896
making shallow model resilient to corrupted .npz files

Julian Kates-Harbeck
authored and
Julian Kates-Harbeck
committed
82935e4
removed necessity for keeping track of shot end cutting

Julian Kates-Harbeck
authored and
Julian Kates-Harbeck
committed
d97f271