From 96c701070e777ed8b3d94014f684cf2f242b64b2 Mon Sep 17 00:00:00 2001
From: Daniel Chin <daniel.chin@nyu.edu>
Date: Wed, 18 Oct 2023 14:56:48 +0400
Subject: [PATCH 1/3] support loading models to CPU

---
 utils/model.py | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/utils/model.py b/utils/model.py
index 45e1f41431..9efea47ad7 100644
--- a/utils/model.py
+++ b/utils/model.py
@@ -7,6 +7,7 @@
 import hifigan
 from model import FastSpeech2, ScheduledOptim
 
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 
 def get_model(args, configs, device, train=False):
     (preprocess_config, model_config, train_config) = configs
@@ -17,7 +18,7 @@ def get_model(args, configs, device, train=False):
             train_config["path"]["ckpt_path"],
             "{}.pth.tar".format(args.restore_step),
         )
-        ckpt = torch.load(ckpt_path)
+        ckpt = torch.load(ckpt_path, map_location=device)
         model.load_state_dict(ckpt["model"])
 
     if train:
@@ -60,7 +61,7 @@ def get_vocoder(config, device):
         config = hifigan.AttrDict(config)
         vocoder = hifigan.Generator(config)
         if speaker == "LJSpeech":
-            ckpt = torch.load("hifigan/generator_LJSpeech.pth.tar")
+            ckpt = torch.load("hifigan/generator_LJSpeech.pth.tar", map_location=device)
         elif speaker == "universal":
             ckpt = torch.load("hifigan/generator_universal.pth.tar")
         vocoder.load_state_dict(ckpt["generator"])

From d57e142e53f8b4e75a3d05b3e9e964c428f70b9f Mon Sep 17 00:00:00 2001
From: Daniel Chin <daniel.chin@nyu.edu>
Date: Wed, 18 Oct 2023 14:59:16 +0400
Subject: [PATCH 2/3] add option to turn off plotting spectrogram

---
 synthesize.py  |  6 +++++-
 utils/tools.py | 25 +++++++++++++++----------
 2 files changed, 20 insertions(+), 11 deletions(-)

diff --git a/synthesize.py b/synthesize.py
index 59a682aa7d..d68cf8559c 100644
--- a/synthesize.py
+++ b/synthesize.py
@@ -84,7 +84,10 @@ def preprocess_mandarin(text, preprocess_config):
     return np.array(sequence)
 
 
-def synthesize(model, step, configs, vocoder, batchs, control_values):
+def synthesize(
+    model, step, configs, vocoder, batchs, control_values, 
+    do_plot_spectrogram=True, 
+):
     preprocess_config, model_config, train_config = configs
     pitch_control, energy_control, duration_control = control_values
 
@@ -105,6 +108,7 @@ def synthesize(model, step, configs, vocoder, batchs, control_values):
                 model_config,
                 preprocess_config,
                 train_config["path"]["result_path"],
+                do_plot_spectrogram, 
             )
 
 
diff --git a/utils/tools.py b/utils/tools.py
index f897430e69..3d7a5c93dd 100644
--- a/utils/tools.py
+++ b/utils/tools.py
@@ -161,7 +161,11 @@ def synth_one_sample(targets, predictions, vocoder, model_config, preprocess_con
     return fig, wav_reconstruction, wav_prediction, basename
 
 
-def synth_samples(targets, predictions, vocoder, model_config, preprocess_config, path):
+def synth_samples(
+    targets, predictions, vocoder, model_config, 
+    preprocess_config, path, 
+    do_plot_spectrogram=True, 
+):
 
     basenames = targets[0]
     for i in range(len(predictions[0])):
@@ -187,15 +191,16 @@ def synth_samples(targets, predictions, vocoder, model_config, preprocess_config
             stats = json.load(f)
             stats = stats["pitch"] + stats["energy"][:2]
 
-        fig = plot_mel(
-            [
-                (mel_prediction.cpu().numpy(), pitch, energy),
-            ],
-            stats,
-            ["Synthetized Spectrogram"],
-        )
-        plt.savefig(os.path.join(path, "{}.png".format(basename)))
-        plt.close()
+        if do_plot_spectrogram:
+            fig = plot_mel(
+                [
+                    (mel_prediction.cpu().numpy(), pitch, energy),
+                ],
+                stats,
+                ["Synthetized Spectrogram"],
+            )
+            plt.savefig(os.path.join(path, "{}.png".format(basename)))
+            plt.close()
 
     from .model import vocoder_infer
 

From 7220cb9a9073f407fffe7a53602cb58d3a496d68 Mon Sep 17 00:00:00 2001
From: Daniel Chin <daniel.chin@nyu.edu>
Date: Wed, 18 Oct 2023 15:09:30 +0400
Subject: [PATCH 3/3] legalize output filename in case text contains
 punctuations

---
 utils/tools.py | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/utils/tools.py b/utils/tools.py
index 3d7a5c93dd..3d0d7eb33c 100644
--- a/utils/tools.py
+++ b/utils/tools.py
@@ -160,6 +160,8 @@ def synth_one_sample(targets, predictions, vocoder, model_config, preprocess_con
 
     return fig, wav_reconstruction, wav_prediction, basename
 
+def legalize_filename(filename: str):
+    return ''.join([x if x.isalnum() else '' for x in filename])[:30]
 
 def synth_samples(
     targets, predictions, vocoder, model_config, 
@@ -199,7 +201,7 @@ def synth_samples(
                 stats,
                 ["Synthetized Spectrogram"],
             )
-            plt.savefig(os.path.join(path, "{}.png".format(basename)))
+            plt.savefig(os.path.join(path, "{}.png".format(legalize_filename(basename))))
             plt.close()
 
     from .model import vocoder_infer
@@ -212,7 +214,8 @@ def synth_samples(
 
     sampling_rate = preprocess_config["preprocessing"]["audio"]["sampling_rate"]
     for wav, basename in zip(wav_predictions, basenames):
-        wavfile.write(os.path.join(path, "{}.wav".format(basename)), sampling_rate, wav)
+        wavfile.write(os.path.join(
+            path, "{}.wav".format(legalize_filename(basename))), sampling_rate, wav)
 
 
 def plot_mel(data, stats, titles):