issues
search
nf-core
/
deepmodeloptim
Stochastic Testing and Input Manipulation for Unbiased Learning Systems
https://nf-co.re/deepmodeloptim
MIT License
23
stars
9
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
[feat] force raytune to have maximum running time
#159
suzannejin
closed
4 months ago
1
[feat] robust way of defining and importing the model class
#158
suzannejin
closed
1 month ago
1
ray init does not automaticly detect GPUs
#157
alessiovignoli
opened
5 months ago
1
deadlock and first nf-test
#156
alessiovignoli
closed
4 months ago
0
[seqera platform] on seqera platform, pipeline crashes since it is allowed just 1.GB of memory by default
#155
mathysgrapotte
closed
4 months ago
1
[gpu] add gpu support in yaml config and raytune_learner
#154
mathysgrapotte
closed
5 months ago
0
Nf test pipeline
#153
alessiovignoli
closed
6 months ago
1
Split correct numtimes
#152
alessiovignoli
closed
6 months ago
1
[feat] add a check on split parameters values.
#151
alessiovignoli
closed
1 month ago
1
Micromamba ps image
#150
alessiovignoli
closed
6 months ago
0
Make pipeline run on GPU (+checks)
#149
luisas
closed
5 months ago
1
[bug] If there is only one metric, analysis step crashes
#148
luisas
closed
1 month ago
1
Performance: define metrics per class
#147
luisas
closed
1 month ago
1
[feat] have test data in a different repo
#146
luisas
closed
1 month ago
2
[feat] transform the entire data
#145
suzannejin
closed
6 months ago
0
[feat] resume ray tune
#144
suzannejin
closed
1 month ago
1
[fix] ray tune runs forever sometimes
#143
suzannejin
closed
4 months ago
3
[feat] add an early stopper for ray tune
#142
suzannejin
closed
1 month ago
2
Add `ps` to all containers
#141
evanfloden
closed
6 months ago
2
Motplotlib to image
#140
alessiovignoli
closed
6 months ago
0
[fix] we want to transform the entire data
#139
suzannejin
closed
6 months ago
0
[feature] add logging functionalities
#138
mathysgrapotte
opened
6 months ago
0
[fix] the output of the forward step should be tensor, never dict
#137
suzannejin
closed
6 months ago
0
[analysis] provide a report of auc vs data modification
#136
suzannejin
closed
1 month ago
1
[test] write unittest for analysis default block
#135
suzannejin
closed
1 month ago
1
[feat] add noise to class labels
#134
suzannejin
closed
1 month ago
1
[Docker] add matplotlib
#133
suzannejin
closed
6 months ago
0
[feature] added analysis default block
#132
suzannejin
closed
6 months ago
1
Batch size too big
#131
alessiovignoli
closed
6 months ago
0
[fix] save config files without serialized objects when ray tune finish
#130
suzannejin
closed
6 months ago
0
[test] Write tests for Predict class
#129
luisas
closed
1 month ago
1
[feat] add padding/truncation as data modification
#128
luisas
closed
1 month ago
1
[feat] add PredictWrapper to do forward pass and get the performance metrics
#127
suzannejin
closed
6 months ago
0
Incorporate check model
#126
alessiovignoli
closed
6 months ago
0
[fix] DNA Experiment test is broken - needs to be fixed and added to the common tests
#125
luisas
closed
4 months ago
0
If batch size to test is bigger than data size, it throws a not easy to understand error. Do a check or a warning.
#124
luisas
closed
1 month ago
1
[fix] solve titanic experiment to be able to add noise to data
#123
suzannejin
closed
2 months ago
0
ray tune related stuff work properly
#122
suzannejin
closed
6 months ago
2
Config yaml breaks (with no good error message) if data_params is not given.
#121
luisas
closed
1 month ago
0
[feat] implement a ray tune results parser class
#120
suzannejin
closed
6 months ago
0
[feat] tune config yaml should be able to take a simple value too
#119
suzannejin
closed
3 months ago
1
Shift to data transform
#118
alessiovignoli
closed
6 months ago
0
[feat] take as input Yaml for transform and split information
#117
alessiovignoli
closed
1 month ago
1
fix bug in shuffle code
#116
suzannejin
closed
6 months ago
0
Add pipeline level ci
#115
JoseEspinosa
closed
6 months ago
1
[bug] FloatRankEncoder and IntRankEncoder are currently applying a standardisation between range [0,1] and not really ranking the data
#114
suzannejin
closed
1 month ago
1
[feat] ray tune report validation accuracy and alternative metrics
#113
suzannejin
closed
6 months ago
0
[fix] put tune config `optimizer` into `optimizer_params`
#112
suzannejin
closed
6 months ago
1
[feat] create a class to handle ray tune results
#111
suzannejin
closed
6 months ago
0
[feat] Add data clipping as possible data modification
#110
luisas
closed
2 months ago
0
Previous
Next