import torch import sys from transformers import AutoTokenizer, AutoModelForMaskedLM import os #如果D:\pyprojs\Bert-VITS2\bert\chinese-roberta-wwm-ext-large\pytorch_model存在就用这个 local_bert = False if os.path.exists("./bert/chinese-roberta-wwm-ext-large/pytorch_model.bin"): local_bert = True tokenizer = AutoTokenizer.from_pretrained("./bert/chinese-roberta-wwm-ext-large") if local_bert else AutoTokenizer.from_pretrained("hfl/chinese-roberta-wwm-ext-large") models = dict() def get_bert_feature(text, word2ph, device=None): if ( sys.platform == "darwin" and torch.backends.mps.is_available() and device == "cpu" ): device = "mps" if not device: device = "cuda" if device not in models.keys(): models[device] = AutoModelForMaskedLM.from_pretrained( "./bert/chinese-roberta-wwm-ext-large" ).to(device) if local_bert else AutoModelForMaskedLM.from_pretrained( "hfl/chinese-roberta-wwm-ext-large" ).to(device) with torch.no_grad(): inputs = tokenizer(text, return_tensors="pt") for i in inputs: inputs[i] = inputs[i].to(device) res = models[device](**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__": 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])