DL-Art-School/codes/trainer/eval
James Betker ee8ceed6da rework tfd13 further
- use a gated activation layer for both attention & convs
- add a relativistic learned position bias. I believe this is similar to the T5 position encodings but it is simpler and learned
- get rid of prepending to the attention matrix - this doesn't really work that well. the model eventually learns to attend one of its heads to these blocks but why not just concat if it is doing that?
2022-07-20 23:28:29 -06:00
..
__init__.py Add sr_fid evaluator 2020-12-30 20:18:58 -07:00
audio_diffusion_fid.py adf for ar-latent tfd 2022-06-15 16:41:08 -06:00
eval_wer.py f 2022-05-27 13:47:05 -06:00
evaluator.py Another fix 2021-06-14 09:51:44 -06:00
fid.py Another fix 2021-06-14 09:51:44 -06:00
flow_gaussian_nll.py more arbitrary fixes 2022-03-17 17:45:44 -06:00
mel_evaluator.py Refactor audio-style models into the audio folder 2022-03-15 11:06:25 -06:00
music_diffusion_fid.py rework tfd13 further 2022-07-20 23:28:29 -06:00
single_point_pair_contrastive_eval.py more arbitrary fixes 2022-03-17 17:45:44 -06:00
sr_diffusion_fid.py Fix a bug where non-rank-0 is computing FID before all images are saved. 2021-06-16 16:27:09 -06:00
sr_fid.py Another fix 2021-06-14 09:51:44 -06:00
sr_style.py More cleaning 2022-03-16 12:05:56 -06:00