joan126
joan126
> `tf.expand_dims(tensor, axis=1)` suggests that the unsqueeze axis should be a constant `1`. > > What TF2ONNX version are you using? Are you able to share the model? tf.expand_dims(tensor, axis=1)...
> Hey @Rayhane-mamah and @begeekmyfriend! Have you tried to fine tune pretrained model on different voice? How much data did you use for it? How much steps did you train...
> https://www.guiji.ai/site/article?id=251 这文章中提到freeze decoder的参数做说话人自适应， 应该是freeze encoder吧...encoder只与文本向量有关，与说话人无关的。
guided attention loss use T,N should represent max_text_length and max_mel_frames of overall dataset or of one batch?
> No it is the average value of all samples. You have to estimate it ahead. In espnet [espnet](https://github.com/espnet/espnet/blob/c370ab2daeabe1d4ceb1747a795b491d3a412936/espnet/nets/pytorch_backend/e2e_tts_tacotron2.py#L25) and moliza/TTS repo, N,T represent max_text_length and max_mel_frame of a batch
> Neither, it just uses the GMM-HMM pipeline for training through LDA+SAT. The experiments that we did with nnet2 several years had very slight benefits if at all over the...
met same issue when run examples/language/gpt/gemini/run_gemini.sh , have you solved this ? @bestbzw
> 不报这个错了，，但是还是无法运行 socket 问题怎么解决的？我也遇到了
> 我也遇到了 之前有人说配置/etc/hosts 配置后，起作用了吗？我配置了也不行，不管是用torchrun还是colossalai run都不行