From 10b00b4da7d2d79418527417af8cc8bf9ac72cbc Mon Sep 17 00:00:00 2001
From: lizi <49679880@qq.com>
Date: Fri, 29 Apr 2022 16:19:26 +0800
Subject: [PATCH 1/2] fix the reorganize_aishell3 trouble now it can generate
 lab files of audio files under training classification

---
 .../other/mfa/local/reorganize_aishell3.py    | 32 +++++++++----------
 1 file changed, 16 insertions(+), 16 deletions(-)

diff --git a/examples/other/mfa/local/reorganize_aishell3.py b/examples/other/mfa/local/reorganize_aishell3.py
index 0ad306626..a97ee29ed 100644
--- a/examples/other/mfa/local/reorganize_aishell3.py
+++ b/examples/other/mfa/local/reorganize_aishell3.py
@@ -46,22 +46,22 @@ def write_lab(root_dir: Union[str, Path],
         text_path = root_dir / sub_set / 'content.txt'
         new_dir = output_dir / sub_set
 
-    with open(text_path, 'r') as rf:
-        for line in rf:
-            wav_id, context = line.strip().split('\t')
-            spk_id = wav_id[:7]
-            transcript_name = wav_id.split('.')[0] + '.lab'
-            transcript_path = new_dir / spk_id / transcript_name
-            context_list = context.split()
-            word_list = context_list[0:-1:2]
-            pinyin_list = context_list[1::2]
-            wf = open(transcript_path, 'w')
-            if script_type == 'word':
-                # add space between chinese char
-                new_context = ' '.join(word_list)
-            elif script_type == 'pinyin':
-                new_context = ' '.join(pinyin_list)
-            wf.write(new_context + '\n')
+        with open(text_path, 'r') as rf:
+            for line in rf:
+                wav_id, context = line.strip().split('\t')
+                spk_id = wav_id[:7]
+                transcript_name = wav_id.split('.')[0] + '.lab'
+                transcript_path = new_dir / spk_id / transcript_name
+                context_list = context.split()
+                word_list = context_list[0:-1:2]
+                pinyin_list = context_list[1::2]
+                wf = open(transcript_path, 'w')
+                if script_type == 'word':
+                    # add space between chinese char
+                    new_context = ' '.join(word_list)
+                elif script_type == 'pinyin':
+                    new_context = ' '.join(pinyin_list)
+                wf.write(new_context + '\n')
 
 
 def reorganize_aishell3(root_dir: Union[str, Path],

From fba0693a208a2818ffea1a874562fe67733071f4 Mon Sep 17 00:00:00 2001
From: Jerryuhoo <jerryuhoo@gmail.com>
Date: Fri, 29 Apr 2022 17:43:50 +0800
Subject: [PATCH 2/2] fix random speaker embedding bug, test=tts

---
 paddlespeech/t2s/exps/voice_cloning.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/paddlespeech/t2s/exps/voice_cloning.py b/paddlespeech/t2s/exps/voice_cloning.py
index 9257b07de..2742cd068 100644
--- a/paddlespeech/t2s/exps/voice_cloning.py
+++ b/paddlespeech/t2s/exps/voice_cloning.py
@@ -110,10 +110,10 @@ def voice_cloning(args):
         print(f"{utt_id} done!")
     # Randomly generate numbers of 0 ~ 0.2, 256 is the dim of spk_emb
     random_spk_emb = np.random.rand(256) * 0.2
-    random_spk_emb = paddle.to_tensor(random_spk_emb)
+    random_spk_emb = paddle.to_tensor(random_spk_emb, dtype='float32')
     utt_id = "random_spk_emb"
     with paddle.no_grad():
-        wav = voc_inference(am_inference(phone_ids, spk_emb=spk_emb))
+        wav = voc_inference(am_inference(phone_ids, spk_emb=random_spk_emb))
     sf.write(
         str(output_dir / (utt_id + ".wav")),
         wav.numpy(),