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

update

Browse Source
This commit is contained in:
游雁 2024-07-08 17:52:38 +08:00
parent d8d1cda2ee
commit 3fe8ff8f8c
1 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
funasr/models/llm_asr/model.py
View File

@ -1530,10 +1530,10 @@ class LLMASR5(nn.Module):
self.concat_emb_hidden = audio_decoder_conf.get("concat_emb_hidden", False)
self.concat_emb_hidden_norm = audio_decoder_conf.get("concat_emb_hidden_norm", False)
if self.concat_emb_hidden_norm:
self.hidden_norm = torch.nn.LayerNorm(self.llm_dim)
self.hidden_norm = torch.nn.LayerNorm(llm_dim)
self.fusion_dropout = nn.Dropout(audio_decoder_conf.get("fusion_drop_rate", 0.0))
self.emb_norm = torch.nn.LayerNorm(self.llm_dim)
self.fusion_norm = torch.nn.LayerNorm(self.llm_dim)
self.emb_norm = torch.nn.LayerNorm(llm_dim)
self.fusion_norm = torch.nn.LayerNorm(llm_dim)
self.fusion_act = Swish()
self.codebook_dim = audio_decoder_conf.get("codebook_dim", 1024)
self.codebook_size = audio_decoder_conf.get("codebook_size", 4096)