From e022acfdea54fd7a6aefb7a353cd5b10a39d3bfd Mon Sep 17 00:00:00 2001
From: 游雁 <zhifu.gzf@alibaba-inc.com>
Date: 星期四, 29 二月 2024 17:23:30 +0800
Subject: [PATCH] v1.0.11
---
funasr/models/seaco_paraformer/model.py | 13 +++----------
1 files changed, 3 insertions(+), 10 deletions(-)
diff --git a/funasr/models/seaco_paraformer/model.py b/funasr/models/seaco_paraformer/model.py
index 0287f56..21ad874 100644
--- a/funasr/models/seaco_paraformer/model.py
+++ b/funasr/models/seaco_paraformer/model.py
@@ -25,8 +25,8 @@
from funasr.train_utils.device_funcs import force_gatherable
from funasr.models.bicif_paraformer.model import BiCifParaformer
from funasr.losses.label_smoothing_loss import LabelSmoothingLoss
-from funasr.utils.timestamp_tools import ts_prediction_lfr6_standard
from funasr.models.transformer.utils.add_sos_eos import add_sos_eos
+from funasr.utils.timestamp_tools import ts_prediction_lfr6_standard
from funasr.models.transformer.utils.nets_utils import make_pad_mask, pad_list
from funasr.utils.load_utils import load_audio_text_image_video, extract_fbank
@@ -223,12 +223,8 @@
# ASF Core
if nfilter > 0 and nfilter < num_hot_word:
- for dec in self.seaco_decoder.decoders:
- dec.reserve_attn = True
- # cif_attended, _ = self.decoder2(contextual_info, _contextual_length, sematic_embeds, ys_pad_lens)
- dec_attended, _ = self.seaco_decoder(contextual_info, _contextual_length, decoder_hidden, ys_pad_lens)
- # cif_filter = torch.topk(self.decoder2.decoders[-1].attn_mat[0][0].sum(0).sum(0)[:-1], min(nfilter, num_hot_word-1))[1].tolist()
- hotword_scores = self.seaco_decoder.decoders[-1].attn_mat[0][0].sum(0).sum(0)[:-1]
+ hotword_scores = self.seaco_decoder.forward_asf6(contextual_info, _contextual_length, decoder_hidden, ys_pad_lens)
+ hotword_scores = hotword_scores[0].sum(0).sum(0)
# hotword_scores /= torch.sqrt(torch.tensor(hw_lengths)[:-1].float()).to(hotword_scores.device)
dec_filter = torch.topk(hotword_scores, min(nfilter, num_hot_word-1))[1].tolist()
add_filter = dec_filter
@@ -239,9 +235,6 @@
contextual_info = selected.squeeze(0).repeat(encoder_out.shape[0], 1, 1).to(encoder_out.device)
num_hot_word = contextual_info.shape[1]
_contextual_length = torch.Tensor([num_hot_word]).int().repeat(encoder_out.shape[0]).to(encoder_out.device)
- for dec in self.seaco_decoder.decoders:
- dec.attn_mat = []
- dec.reserve_attn = False
# SeACo Core
cif_attended, _ = self.seaco_decoder(contextual_info, _contextual_length, sematic_embeds, ys_pad_lens)
--
Gitblit v1.9.1