Releases: albertaparicio/tfglib
Releases · albertaparicio/tfglib
v1.3.0
v1.2.1: Pretrain generator: output sequence batches, instead of individual se…
…quences Bump version number up to 1.2.1
v1.2.0
Bump version number to 1.2
v1.1.0
Change log (by listing the commits in between v1.0.0 and v1.1.0):
- e112a04 Change encoder-decoder option parameter for an offset parameter
- 8c392aa Fix seq2seq 2-model weight loading
- d0380bb Typo fix in backend import
- 67bbb82 Add s2s_load_weights
- fb1b0d7 Add display_time method
- 455805c Unscale prediction and mask data methods
- 2566121 Minmax scale both source and trained masked data
- c9378aa Add minmax scaling function
- 4c069e0 Compute maximum and minimum values of each speaker's parameters
- f172a93 Fix zero-padding and mask construction
- f72c43c Save and load max_seq_length as an attribute of h5py
- fef9fcb Fix issue with max_seq_length extraction
- 30739e7 Reshape masks into 2D matrices
- afbe1c1 Apply compression to dataset after .h5 saving
- aee82fd Rename seq2seq_datatable and remove printing of speakers and basenames
- 161e228 Add zero-padding modes and minor fix in h5 saving
- f707567 Add zero_pad_params method
v1.0.0: Version 1.0.0
First version installable with pip