Solved the problem that the existing model could not be loaded when training the GAN model (#549)

* The new vocoder Fre-GAN is now supported

* Improved some fregan details

* Fixed the problem that the existing model could not be loaded to continue training when training GAN

* Updated reference papers

README-CN.md

@@ -141,6 +141,7 @@
 | --- | ----------- | ----- | --------------------- |
 | [1803.09017](https://arxiv.org/abs/1803.09017) | GlobalStyleToken (synthesizer)| Style Tokens: Unsupervised Style Modeling, Control and Transfer in End-to-End Speech Synthesis | 本代码库 |
 | [2010.05646](https://arxiv.org/abs/2010.05646) | HiFi-GAN (vocoder)| Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis | 本代码库 |
+| [2106.02297](https://arxiv.org/abs/2106.02297) | Fre-GAN (vocoder)| Fre-GAN: Adversarial Frequency-consistent Audio Synthesis | 本代码库 |
 |[**1806.04558**](https://arxiv.org/pdf/1806.04558.pdf) | SV2TTS | Transfer Learning from Speaker Verification to Multispeaker Text-To-Speech Synthesis | 本代码库 |
 |[1802.08435](https://arxiv.org/pdf/1802.08435.pdf) | WaveRNN (vocoder) | Efficient Neural Audio Synthesis | [fatchord/WaveRNN](https://github.com/fatchord/WaveRNN) |
 |[1703.10135](https://arxiv.org/pdf/1703.10135.pdf) | Tacotron (synthesizer) | Tacotron: Towards End-to-End Speech Synthesis | [fatchord/WaveRNN](https://github.com/fatchord/WaveRNN)

README.md

@@ -97,6 +97,7 @@ You can then try the toolbox:
 | --- | ----------- | ----- | --------------------- |
 | [1803.09017](https://arxiv.org/abs/1803.09017) | GlobalStyleToken (synthesizer)| Style Tokens: Unsupervised Style Modeling, Control and Transfer in End-to-End Speech Synthesis | This repo |
 | [2010.05646](https://arxiv.org/abs/2010.05646) | HiFi-GAN (vocoder)| Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis | This repo |
+| [2106.02297](https://arxiv.org/abs/2106.02297) | Fre-GAN (vocoder)| Fre-GAN: Adversarial Frequency-consistent Audio Synthesis | This repo |
 |[**1806.04558**](https://arxiv.org/pdf/1806.04558.pdf) | **SV2TTS** | **Transfer Learning from Speaker Verification to Multispeaker Text-To-Speech Synthesis** | This repo |
 |[1802.08435](https://arxiv.org/pdf/1802.08435.pdf) | WaveRNN (vocoder) | Efficient Neural Audio Synthesis | [fatchord/WaveRNN](https://github.com/fatchord/WaveRNN) |
 |[1703.10135](https://arxiv.org/pdf/1703.10135.pdf) | Tacotron (synthesizer) | Tacotron: Towards End-to-End Speech Synthesis | [fatchord/WaveRNN](https://github.com/fatchord/WaveRNN)

vocoder/fregan/train.py

@@ -51,8 +51,8 @@ def train(rank, a, h):
         print("checkpoints directory : ", a.checkpoint_path)
 
     if os.path.isdir(a.checkpoint_path):
-        cp_g = scan_checkpoint(a.checkpoint_path, 'g_')
+        cp_g = scan_checkpoint(a.checkpoint_path, 'g_fregan_')
-        cp_do = scan_checkpoint(a.checkpoint_path, 'do_')
+        cp_do = scan_checkpoint(a.checkpoint_path, 'do_fregan_')
 
     steps = 0
     if cp_g is None or cp_do is None:

vocoder/fregan/utils.py

@@ -58,7 +58,7 @@ def save_checkpoint(filepath, obj):
 
 
 def scan_checkpoint(cp_dir, prefix):
-    pattern = os.path.join(cp_dir, prefix + '????????')
+    pattern = os.path.join(cp_dir, prefix + '????????.pt')
     cp_list = glob.glob(pattern)
     if len(cp_list) == 0:
         return None

vocoder/hifigan/train.py

@@ -51,8 +51,8 @@ def train(rank, a, h):
         print("checkpoints directory : ", a.checkpoint_path)
 
     if os.path.isdir(a.checkpoint_path):
-        cp_g = scan_checkpoint(a.checkpoint_path, 'g_')
+        cp_g = scan_checkpoint(a.checkpoint_path, 'g_hifigan_')
-        cp_do = scan_checkpoint(a.checkpoint_path, 'do_')
+        cp_do = scan_checkpoint(a.checkpoint_path, 'do_hifigan_')
 
     steps = 0
     if cp_g is None or cp_do is None:
@@ -181,10 +181,10 @@ def train(rank, a, h):
 
                 # checkpointing
                 if steps % a.checkpoint_interval == 0 and steps != 0:
-                    checkpoint_path = "{}/g_{:08d}.pt".format(a.checkpoint_path, steps)
+                    checkpoint_path = "{}/g_hifigan_{:08d}.pt".format(a.checkpoint_path, steps)
                     save_checkpoint(checkpoint_path,
                                     {'generator': (generator.module if h.num_gpus > 1 else generator).state_dict()})
-                    checkpoint_path = "{}/do_{:08d}.pt".format(a.checkpoint_path, steps)
+                    checkpoint_path = "{}/do_hifigan_{:08d}.pt".format(a.checkpoint_path, steps)
                     save_checkpoint(checkpoint_path,
                                     {'mpd': (mpd.module if h.num_gpus > 1 else mpd).state_dict(),
                                      'msd': (msd.module if h.num_gpus > 1 else msd).state_dict(),

vocoder/hifigan/utils.py

@@ -50,7 +50,7 @@ def save_checkpoint(filepath, obj):
 
 
 def scan_checkpoint(cp_dir, prefix):
-    pattern = os.path.join(cp_dir, prefix + 'hifigan.pt')
+    pattern = os.path.join(cp_dir, prefix + '????????.pt')
     cp_list = glob.glob(pattern)
     if len(cp_list) == 0:
         return None