doheeeeeee

batchsize should be 42

I know about SyncBatchNorm.But DeepInversion should calculate loss about each pixel and my gpu can't bear it.

@hdmjdp I can't understand prefix. What does it mean? Here is the shell script i ran. I just changed the "num-epochs", "max-duration" and "world-size". ./run.sh --stage 4 --stop-stage 4 --max-duration...

@hdmjdp I ran [vall-e last week version](https://github.com/lifeiteng/vall-e/tree/93c7d5923d4eb5e468a65bf1d0a5df7dfbe643d8) which has no prefix option. And I found prefix 0 is same as [vall-e last week version](https://github.com/lifeiteng/vall-e/tree/93c7d5923d4eb5e468a65bf1d0a5df7dfbe643d8) version. Here is my tensorboard image....

@hdmjdp Here is my tensorboard log. [tensorboard](https://tensorboard.dev/experiment/AFBURwy2SnqPiNC40GBAfQ/#scalars)

@hdmjdp The prompt speakers are in test-clean not training data.

> based on the latest commit? Thanks based on [last week commit](https://github.com/lifeiteng/vall-e/tree/93c7d5923d4eb5e468a65bf1d0a5df7dfbe643d8). thank you

@thangnvkcn @jieen1 @LorenzoBrugioni @UncleSens @Zhang-Xiaoyi @lqj01 @UESTCgan @yiwei0730 @hackerxiaobai Sorry for late reply. This is the model that I trained. google drive link : [link](https://drive.google.com/file/d/1pKvS56NnzVCYqhbjoFqL_JQ8i32UixbL/view?usp=sharing) infer like this command: `python...

Hello. I trained CR-CTC model and decoded streaming CTC model and got token repetition (ex. ref: 안녕하세요 / hyp: 안녕녕하세요) So, I really need online prefix beam search.... Do you...

Hello, when will you plan to support CTC streaming decoding? I checked sherpa-onnx but I can't find sherpa (torchscript model) support. Thank you.