tuna2134/sbv2-v2
1
0
Fork 0
Code Issues Pull Requests 1 Actions Packages Projects Releases Wiki Activity

Apply Code Formatter Change

Browse Source
This commit is contained in:
Stardust-minus
2023-09-06 13:42:05 +00:00
committed by github-actions[bot]
parent d82ba3457a
commit eef212fd00
26 changed files with 2938 additions and 1453 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
text/japanese_bert.py
View File

@@ -4,19 +4,26 @@ import sys
tokenizer = AutoTokenizer.from_pretrained("./bert/bert-base-japanese-v3")
def get_bert_feature(text, word2ph, device=None):
if sys.platform == "darwin" and torch.backends.mps.is_available() and device == "cpu":
if (
sys.platform == "darwin"
and torch.backends.mps.is_available()
and device == "cpu"
):
device = "mps"
if not device:
device = "cuda"
model = AutoModelForMaskedLM.from_pretrained("./bert/bert-base-japanese-v3").to(device)
model = AutoModelForMaskedLM.from_pretrained("./bert/bert-base-japanese-v3").to(
device
)
with torch.no_grad():
inputs = tokenizer(text, return_tensors='pt')
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 inputs['input_ids'].shape[-1] == len(word2ph)
res = torch.cat(res["hidden_states"][-3:-2], -1)[0].cpu()
assert inputs["input_ids"].shape[-1] == len(word2ph)
word2phone = word2ph
phone_level_feature = []
for i in range(len(word2phone)):