-
Notifications
You must be signed in to change notification settings - Fork 1
/
CHANGE.txt
32 lines (27 loc) · 1.2 KB
/
CHANGE.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
v0.0.7:
1. Bugfix: update dependency version of `longling` to solve the problem using torch light_module with cuda
v0.0.6:
1. Bugfix: update dependency version of `longling`
v0.0.5:
1. mxnet: add split_and_load
2. add bucketing
v0.0.4:
1. utils: add io relevant including json_load, as_out_io, as_io
2. utils: add path relevant including path_append
3. torch: add rnn layer: get_state_shape, get_init_state
4. torch: add erase-add gate
5. torch: add functional.utils including tensor2list, batch_select
6. torch: add functional.seq including pick, length2mask, get_sequence_mask, mask_sequence
7. torch: add functional.rnn including get_rnn_state_shape, get_rnn_init_state
8. add analysis relevant: get_best
9. add get_epoch_params_filepath
10. add Timer, as_list, config_logging, iterwrap
11. update dependencies: longling
12. add constants: CFG_JSON, RESULT_JSON
13. add new function: get_params_filepath
v0.0.3:
1. add configuration parser
2. add save_params and load_net for torch and mxnet
v0.0.2:
1. Add light_module for both torch and mxnet
2. add modelzoo for mxnet, including attention, highway, pairwise loss and TextCNN and TransE