Junity

Question about KL Divergence loss function 关于KL散度的损失函数的问题

7

English and Chinese are translations to each other.  This is loss function for KL Divergence. Specifically, for these formulas: 以下是KL散度的损失函数，特别是这段公式：  they are acutally calculating kl divergence for gaussian...

Add support for train without specify pretrained model, add support for selecting v2 48k as training setting, and add support for auto removing pretrained model path in corresponding box when...

Stop extracting features when hubert_base.pt does not exist.

3

When there is no hubert_base.pt, the training process will not stop, and it will normally run through each epoch. However, the speed is abnormally fast, and the result is a...

Support choosing speech encoder(feature extractor)

2

I add a new option, PPGLarge. We may also add more option of speech encoder if there is one that perform the best. In addition, I add a projection in...

Support half precision for VQ and FSQ

6

Hi, I notice it is not support other preicsion. I made this tiny change and I tried a simple sample and it works on fp16 and bf16. I noticed that...

Dear all, Thank you for your amazing works. I would like to build an arena that for specific tasks. Could I get the exact code of chat.lmsys.org and I can...

Inference Speed Benchmark 推理速度测评 [COME AND SHARE YOUR SPEED!]

19

I hope that everyone enter this issue can share your system, CPU, GPU, inference speed in GPT stage, and the version you use(better to compare v2 as a standard). so...