GateNLP gate-lf-pytorch-json issues

GateNLP / gate-lf-pytorch-json

PyTorch wrapper for the LearningFramework GATE plugin

Apache License 2.0

1 stars 2 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

Allow usage of pytorch-hub

#49 johann-petrak opened 5 years ago
0
Pytorch emits an odd message which screws up the process communication

#48 johann-petrak closed 5 years ago
6
1905useorig: changes for using Elmo

#47 johann-petrak closed 5 years ago
0
Make sure validation can be done using several batches!

#46 johann-petrak opened 5 years ago
0
Make it possible to use the original format data

#45 johann-petrak closed 5 years ago
8
Check use of ngrams with chunking!

#44 johann-petrak closed 5 years ago
1
Chunking with a numerical feature causes an exception

#43 johann-petrak closed 5 years ago
2
Use same metric for early stopping and saving best model

#42 johann-petrak closed 5 years ago
0
Test if moving the model between machines works

#41 johann-petrak closed 5 years ago
1
Always use CPU for application for now

#40 johann-petrak closed 5 years ago
2
Better way to store the meta file with the modelwrapper or not store at all

#39 johann-petrak closed 5 years ago
4
Make sure nothing on CUDA gets saved to the wrapper file

#38 johann-petrak closed 5 years ago
0
Use the loss function configured from the modelwrapper/module

#37 johann-petrak closed 5 years ago
0
Make logging output more concise

#36 johann-petrak closed 5 years ago
1
Problem at application time on a CUDA-enabled system

#35 johann-petrak closed 5 years ago
6
Setting seed does not allow to exactly reproduce results

#34 johann-petrak closed 5 years ago
3
Allow CNN/LSTM to get chosen for ngrams from the command line

#33 johann-petrak closed 5 years ago
1
Make early stopping parametrizable through a command line option

#32 johann-petrak closed 5 years ago
0
Add module / layer for BiLSTM-CRF

#31 johann-petrak opened 5 years ago
1
Add special module for LSTM-based sentence classification

#30 johann-petrak closed 5 years ago
1
Check correct working with pre-trained embeddings and sentence classification

#29 johann-petrak closed 5 years ago
1
Create our own module API

#28 johann-petrak opened 5 years ago
0
Simplify and adapt with simplified python-data/lf

#27 johann-petrak opened 5 years ago
0
Find a way how to pass arbitrary parameters/configuration from the command line and/or from a file

#26 johann-petrak opened 5 years ago
8
Make sure we use a good way to create the python documentation

#25 johann-petrak closed 5 years ago
1
Test under all supported python versions and environments

#24 johann-petrak opened 6 years ago
0
Figure out the difference between importing a user module in python 3.5 vs. 3.6

#23 johann-petrak closed 6 years ago
1
Allow to include regularization terms in the loss function

#22 johann-petrak opened 6 years ago
0
Allow for gradient clipping

#21 johann-petrak opened 6 years ago
2
Check why training using pre-trained embeddings or mapping does not work

#20 johann-petrak closed 6 years ago
2
Make sure we do not save unnecessary data with the model

#19 johann-petrak opened 6 years ago
2
Move all pre-defined models into their own modelzoo directory

#18 johann-petrak closed 6 years ago
1
Make it easy to specify a model wrapper

#17 johann-petrak opened 6 years ago
0
Make sure python Path methods are compatible with 3.x < 3.6

#16 johann-petrak opened 6 years ago
0
--cuda False does not seem to work

#15 johann-petrak closed 6 years ago
2
When there is a file, only use the sanitized base name in the layer name

#14 johann-petrak opened 6 years ago
1
Proper interaction of save/load and cpu/gpu transfer

#13 johann-petrak opened 6 years ago
0
Upgrade code to 0.4 of pytorch and require 0.4 minimum

#12 johann-petrak opened 6 years ago
2
Check model file size after saving

#11 johann-petrak closed 6 years ago
2
Sequence application needs to get properly implemented!

#10 johann-petrak opened 6 years ago
1
Correct sharing of embedding parameters if same embedding id!

#9 johann-petrak opened 6 years ago
0
Delegate pickling details for the dataset instance to the Dataset class

#8 johann-petrak opened 6 years ago
0
Move handling of pre-calculated embeddings from the vocab to the module

#7 johann-petrak closed 6 years ago
0
Better handling of training/validation set evaluations

#6 johann-petrak opened 6 years ago
0
Separate function for validation set evaluation

#5 johann-petrak opened 6 years ago
0
Make sure evaluation is correctly done for sequences

#4 johann-petrak opened 6 years ago
0
Unique identifier for every layer we create

#3 johann-petrak opened 6 years ago
1
Calculate / override LSTM hidden layer size, override architecture details

#2 johann-petrak opened 6 years ago
1
Consider interfacing to tensorboard or CometML or both

#1 johann-petrak opened 6 years ago
1