init

text/chinese_bert.py

@@ -0,0 +1,55 @@
+import torch
+from transformers import AutoTokenizer, AutoModelForMaskedLM
+import os
+
+os.environ['ALL_PROXY']='socks5://127.0.0.1:7890'
+os.environ['HTTPS_PROXY']='http://127.0.0.1:7890'
+os.environ['HTTP_PROXY']='http://127.0.0.1:7890'
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+tokenizer = AutoTokenizer.from_pretrained("hfl/chinese-roberta-wwm-ext-large")
+model = AutoModelForMaskedLM.from_pretrained("hfl/chinese-roberta-wwm-ext-large").to(device)
+
+def get_bert_feature(text, word2ph):
+    with torch.no_grad():
+        inputs = tokenizer(text, return_tensors='pt')
+        for i in inputs:
+            inputs[i] = inputs[i].to(device)
+        res = model(**inputs, output_hidden_states=True)
+        res = torch.cat(res['hidden_states'][-3:-2], -1)[0].cpu()
+
+    assert len(word2ph) == len(text)+2
+    word2phone = word2ph
+    phone_level_feature = []
+    for i in range(len(word2phone)):
+        repeat_feature = res[i].repeat(word2phone[i], 1)
+        phone_level_feature.append(repeat_feature)
+
+    phone_level_feature = torch.cat(phone_level_feature, dim=0)
+
+
+    return phone_level_feature.T
+
+if __name__ == '__main__':
+    # feature = get_bert_feature('你好,我是说的道理。')
+    import torch
+
+    word_level_feature = torch.rand(38, 1024)  # 12个词,每个词1024维特征
+    word2phone = [1, 2, 1, 2, 2, 1, 2, 2, 1, 2, 2, 1, 2, 2, 2, 2, 2, 1, 1, 2, 2, 1, 2, 2, 2, 2, 1, 2, 2, 2, 2, 2, 1, 2, 2, 2, 2, 1]
+
+    # 计算总帧数
+    total_frames = sum(word2phone)
+    print(word_level_feature.shape)
+    print(word2phone)
+    phone_level_feature = []
+    for i in range(len(word2phone)):
+        print(word_level_feature[i].shape)
+
+        # 对每个词重复word2phone[i]次
+        repeat_feature = word_level_feature[i].repeat(word2phone[i], 1)
+        phone_level_feature.append(repeat_feature)
+
+    phone_level_feature = torch.cat(phone_level_feature, dim=0)
+    print(phone_level_feature.shape)  # torch.Size([36, 1024])
+