speech_recognition/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR synced 2025-09-15 14:48:36 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge pull request #4 from TeaPoly/main

Browse Source 
Fix some issue to make batch inference easy for predictor and decoder.
This commit is contained in:
zhifu gao 2022-12-02 19:23:30 +08:00 committed by GitHub
commit 419533aa3a
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
funasr/bin/asr_inference_paraformer.py
View File

@ -189,7 +189,7 @@ class Speech2Text:
predictor_outs = self.asr_model.calc_predictor(enc, enc_len)
pre_acoustic_embeds, pre_token_length = predictor_outs[0], predictor_outs[1]
pre_token_length = torch.tensor([pre_acoustic_embeds.size(1)], device=pre_acoustic_embeds.device)
pre_token_length = pre_token_length.long()
decoder_outs = self.asr_model.cal_decoder_with_predictor(enc, enc_len, pre_acoustic_embeds, pre_token_length)
decoder_out, ys_pad_lens = decoder_outs[0], decoder_outs[1]

2
funasr/models/predictor/cif.py
View File

@ -147,7 +147,7 @@ class CifPredictorV2(nn.Module):
b, t, d = hidden.size()
tail_threshold = self.tail_threshold
tail_threshold = torch.tensor([tail_threshold], dtype=alphas.dtype).to(alphas.device)
tail_threshold = torch.reshape(tail_threshold, (1, 1))
tail_threshold = tail_threshold.unsqueeze(0).repeat(b, 1)
alphas = torch.cat([alphas, tail_threshold], dim=1)
zeros = torch.zeros((b, 1, d), dtype=hidden.dtype).to(hidden.device)
hidden = torch.cat([hidden, zeros], dim=1)