jeremy110

@smlkdev 1. Basically, fine-tuning usually requires less than 10 epochs; the settings from others are used as a reference.  2. Typically, a model trained for 10 epochs is already...

@kadirnar hello~ Typically, the parameters would be read from the YAML file, but it's okay. You can stop it at the appropriate time. I’ve kind of forgotten how many steps...

@manhcuong17072002 This approach can indeed enhance the data and may provide a slight improvement, but several points need to be considered. Since MeloTTS uses BERT to extract feature vectors, if...

模型架構是一樣的，只是多新增了使用IPA訓練新語言。

@lukeewin 或許你可以參考這個 https://github.com/myshell-ai/MeloTTS/issues/120 主要就是 1. 新增 symbols 2. 替換 bert 3. 取代 g2p

1. 基本上任何架構的 bert 都可以，但要是你的語言訓練出來的，或者拿你的語言去做 fine-tune。 2. g2p 主要也是看你的語言如何將文字轉成拼音，也就是最後 phones 的格式。 在我這邊是轉成拼音，拼音再轉成 ipa，最後用 ipa 去做訓練。

抱歉 沒辦法 最近公司專案有點多，你那個需要花很多時間fine-tune bert，而且不一定會成功，另外 g2p 方面我不是專家，我是直接拿同事的程式來跑的。

@deepanshu-yadav Hi~ Based on my previous fine-tuning experience, here are my responses to the following points: 1. Training a 0.6B model on a 16GB GPU is quite challenging. NeMo typically...

@deepanshu-yadav I would recommend at least a 3090 or 4090, which can train a 110M model, but you'll need at least 1,000 hours of data.

@BakingBrains Hi~ Did you also train with a small amount of data? Here's a method I personally find quite effective: using AdamW8bit. If your machine supports it, it can reduce...