Releases: lucidrains/spear-tts-pytorch
Releases · lucidrains/spear-tts-pytorch
0.4.8
0.4.7
omit the wav2vec parameters from optimizers
0.4.6
make sure wav2vec is done in eval mode and without gradients
0.4.5
address https://github.com/lucidrains/spear-tts-pytorch/pull/9
0.4.4
quick fix for spec decoding
0.4.3
fix default topk sampling filter threshold
0.4.2
Update setup.py
0.4.1
take care of eos during spec decoding
0.4.0
finally complete speculative decoding with early exit strategy for mo…
0.3.14
give the early exit path an extra feedforward transformer block, and …