From a361df3186827d3e84bc30b8e6b41a31c11c50e8 Mon Sep 17 00:00:00 2001
From: erogol <erogol@hotmail.com>
Date: Sun, 31 May 2020 03:31:42 +0200
Subject: [PATCH] plot fix

---
 vocoder/configs/melgan_config.json | 7 +++----
 vocoder/train.py                   | 4 ++--
 vocoder/utils/generic_utils.py     | 8 ++++----
 3 files changed, 9 insertions(+), 10 deletions(-)

diff --git a/vocoder/configs/melgan_config.json b/vocoder/configs/melgan_config.json
index 9a3ded37..7da42f75 100644
--- a/vocoder/configs/melgan_config.json
+++ b/vocoder/configs/melgan_config.json
@@ -109,14 +109,13 @@
     "test_sentences_file": null,  // set a file to load sentences to be used for testing. If it is null then we use default english sentences.
 
     // OPTIMIZER
-    "noam_schedule": true,        // use noam warmup and lr schedule.
-    "grad_clip": 1.0,              // upper limit for gradients for clipping.
+    "noam_schedule": false,        // use noam warmup and lr schedule.
+    "warmup_steps_gen": 4000,          // Noam decay steps to increase the learning rate from 0 to "lr"
+    "warmup_steps_disc": 4000,
     "epochs": 1000,                // total number of epochs to train.
     "wd": 0.000001,                // Weight decay weight.
     "lr_gen": 0.0001,                  // Initial learning rate. If Noam decay is active, maximum learning rate.
     "lr_disc": 0.0001,
-    "warmup_steps_gen": 4000,          // Noam decay steps to increase the learning rate from 0 to "lr"
-    "warmup_steps_disc": 4000,
     "gen_clip_grad": 10.0,
     "disc_clip_grad": 10.0,
 
diff --git a/vocoder/train.py b/vocoder/train.py
index b0b4833a..0d039e73 100644
--- a/vocoder/train.py
+++ b/vocoder/train.py
@@ -372,11 +372,11 @@ def evaluate(model_G, criterion_G, model_D, ap, global_step, epoch):
             c_logger.print_eval_step(num_iter, loss_dict, keep_avg.avg_values)
 
     # compute spectrograms
-    figures = plot_results(y_hat, y_G, ap, global_step, 'eval')
+    figures = plot_results(in_fake_D, in_real_D, ap, global_step, 'eval')
     tb_logger.tb_eval_figures(global_step, figures)
 
     # Sample audio
-    sample_voice = y_hat[0].squeeze(0).detach().cpu().numpy()
+    sample_voice = in_fake_D[0].squeeze(0).detach().cpu().numpy()
     tb_logger.tb_eval_audios(global_step, {'eval/audio': sample_voice},
                              c.audio["sample_rate"])
 
diff --git a/vocoder/utils/generic_utils.py b/vocoder/utils/generic_utils.py
index 062de160..179c4064 100644
--- a/vocoder/utils/generic_utils.py
+++ b/vocoder/utils/generic_utils.py
@@ -13,8 +13,8 @@ def plot_results(y_hat, y, ap, global_step, name_prefix):
     y_hat = y_hat[0].squeeze(0).detach().cpu().numpy()
     y = y[0].squeeze(0).detach().cpu().numpy()
 
-    spec_fake = ap.spectrogram(y_hat).T
-    spec_real = ap.spectrogram(y).T
+    spec_fake = ap.melspectrogram(y_hat).T
+    spec_real = ap.melspectrogram(y).T
     spec_diff = np.abs(spec_fake - spec_real)
 
     # plot figure and save it
@@ -98,5 +98,5 @@ def setup_discriminator(c):
     return model
 
 
-# def check_config(c):
-#     pass
\ No newline at end of file
+def check_config(c):
+    pass
\ No newline at end of file