Age | Commit message (Collapse) | Author | |
---|---|---|---|
2023-09-26 | changed checkpoint formatexp_pitchdnn1 | Jan Buethe | |
2023-09-25 | Adapting to new data format/model | Jean-Marc Valin | |
2023-09-25 | Python code for neural pitch | Krishna Subramani | |
2023-09-25 | Add noise augmentation to pitch dumping | Jean-Marc Valin | |
2023-09-23 | Add pitch feature computation | Jean-Marc Valin | |
2023-09-15 | Band-limiting the voicing parameter | Jean-Marc Valin | |
2023-09-13 | removed trailing whitespace in fargan | Jan Buethe | |
Signed-off-by: Jan Buethe <jbuethe@amazon.de> | |||
2023-09-13 | integrated JM's FFT ada conv | Jan Buethe | |
Signed-off-by: Jan Buethe <jbuethe@amazon.de> | |||
2023-09-13 | Remove --has-gain and --passthrough-size args | Jean-Marc Valin | |
Couldn't be used anymore | |||
2023-09-13 | Add conditioning interpolation, fwconv layer | Jean-Marc Valin | |
2023-09-13 | Only use one frame of pre-loading | Jean-Marc Valin | |
2023-09-13 | Randomly double the training sequence length | Jean-Marc Valin | |
Helps with stability with little cost in training speed | |||
2023-09-13 | Adds skip connections | Jean-Marc Valin | |
2023-09-13 | 5-tap pitch predictor | Jean-Marc Valin | |
2023-09-13 | Also use previous frame | Jean-Marc Valin | |
2023-09-13 | explicit signal gain, explicit pitch predictor | Jean-Marc Valin | |
2023-09-13 | FARGAN initial commit in Opus | Jean-Marc Valin | |
Copied/adapted from LPCNet repo | |||
2023-09-12 | fix | Jan Buethe | |
Signed-off-by: Jan Buethe <jbuethe@amazon.de> | |||
2023-09-12 | added requirements.txt to osce | Jan Buethe | |
Signed-off-by: Jan Buethe <jbuethe@amazon.de> | |||
2023-09-12 | added more enhancement stuff | Jan Buethe | |
Signed-off-by: Jan Buethe <jbuethe@amazon.de> | |||
2023-09-05 | added copyright headers | Jan Buethe | |
Signed-off-by: Jan Buethe <jbuethe@amazon.de> | |||
2023-09-05 | added LPCNet torch implementation | Jan Buethe | |
Signed-off-by: Jan Buethe <jbuethe@amazon.de> | |||
2023-09-05 | brought NoLACE up to date | Jan Buethe | |
Signed-off-by: Jan Buethe <jbuethe@amazon.de> | |||
2023-09-02 | Fixes vnni macro redefinition with clang | Michael Klingbeil | |
2023-09-01 | renamed ShapeNet to NoLACE | Jan Buethe | |
Signed-off-by: Jan Buethe <jbuethe@amazon.de> | |||
2023-08-10 | Changing DRED exp. ID so we can reserve 127 | Jean-Marc Valin | |
There was a suggestion to use 127 for extending the extension ID space | |||
2023-08-10 | Don't allocate DRED bits if we don't have enough | Jean-Marc Valin | |
2023-08-10 | Making it easier to remove DRED experimental ID | Jean-Marc Valin | |
When ready, change DRED_EXTENSION_ID to the final ID, remove DRED_EXPERIMENTAL_VERSION completely, and change DRED_EXPERIMENTAL_BYTES to zero (eventually remove it). | |||
2023-08-08 | improved auto-scaling in wexchange | Jan Buethe | |
2023-08-04 | Add fwgan_load_model() | Jean-Marc Valin | |
2023-08-04 | int version for convenience | Jean-Marc Valin | |
2023-08-04 | Adding some comments | Jean-Marc Valin | |
2023-08-04 | Continuation now working | Jean-Marc Valin | |
Added required filtering/delay/memory to get continuation | |||
2023-08-03 | More scaling/filtering to run_fwgan_subframe() | Jean-Marc Valin | |
2023-08-03 | More FWGAN refactoring | Jean-Marc Valin | |
2023-08-03 | FWGAN refactoring | Jean-Marc Valin | |
2023-08-03 | Applying continuation after first subframe | Jean-Marc Valin | |
Continuation matches Python code now | |||
2023-08-02 | Avoid sin()/cos() calls for pitch reference | Jean-Marc Valin | |
2023-08-02 | Don't compute linear activation in-place | Jean-Marc Valin | |
saves a few cycles | |||
2023-08-02 | Add FWGAN to lpcnet_demoexp-fwgan10 | Jean-Marc Valin | |
2023-08-02 | C implementation of FWGAN | Jean-Marc Valin | |
2023-08-02 | Skeleton for FWGAN code | Jean-Marc Valin | |
2023-08-02 | Handle float matrices with multiple of 8 rows | Jean-Marc Valin | |
2023-08-02 | Add Gated Linear Unit (GLU) | Jean-Marc Valin | |
2023-08-01 | updated dump_model_weights.py | Jan Buethe | |
2023-08-01 | added FWGAN weight dumping code | Jan Buethe | |
2023-08-01 | updadet wexchange version number | Jan Buethe | |
2023-08-01 | added .copy() to weights in wexchange | Jan Buethe | |
2023-08-01 | added auto-scaling to wexchange | Jan Buethe | |
2023-07-29 | fixed wexchange for GRUs without bias | Jan Buethe | |