From 9b46b760b705c4ac278386d14cb6cb7eb1ec6b3d Mon Sep 17 00:00:00 2001
From: dohe0342 <kimdohe1070@gmail.com>
Date: Tue, 11 Apr 2023 13:46:49 +0900
Subject: [PATCH] from local

---
 .../.data2vec_audio.py.swp                    | Bin 36864 -> 40960 bytes
 .../data2vec_audio.py                         |   2 ++
 2 files changed, 2 insertions(+)

diff --git a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.data2vec_audio.py.swp b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.data2vec_audio.py.swp
index bd03ab86d3e1bdf73e5d3716ee11e21a118efce7..3fbe08023686e43031762d1146a59d9ff4c60470 100644
GIT binary patch
delta 916
zcmXxiT}YEr7{KvoU$*w)W*>X4i!q{s)||T#Mi+%yxFiJy&B}t-Hf_vMwwXn+c@eD-
z1ns3P`h+f0i9WWANTIweBAXG>O-8hW>Y}<y2x|RjcHqbRyytM<_nhZ>hHhwwX0*yd
z->#+>*Cww|7<Q45g}QKS^!$D0t4(86b@pM6QysM{70XNOjLv@#d7Zmt57@FUy~|oO
zi>Man+3r*;>Ic(Lcq`3J;4ID{hyz%UPwbe)Zgip+^CpozUSJe$@WYKfe;>yMoPZZD
z2#QA5aS>xk7$q%D1go(Mg=Hci@Dew18Zp$vfQ2%V&&cB*?&2f{(F#9m;K2&~HjpJe
zKnBNf6cOyiRzsMN6Rd-vs27>TFhZ!qdmT5$LtI4+ez>s`^IDN<WYB{Sw4)qWeAnbE
z8zxNZbhSktX))I(`{JQP-cT@E)zuvdCI{k?gvZq%?>pR|++QuOr4j93np4|0JJhv1
z%(-99W!uuhu*l=m0ZIp^i2rL`w2?K;VH~#*Lk)hEi(JPw9Dy4O9C%};97v%T3cgYH
zw|IsD^dNvn3*}M>-j#X)Qy9TMbYUYlU=6;MdI4Ev(2OncLBRt3@DtDR2xCYgi3Ijw
zJ3RP9Ys_NCnCA8bw{Z@qkU<myG@=r(=#@zvM-?0p6zGX8Qs~8AxS_=?8NGumxQq_8
UV=>cgUmQ7$eoB{fHas=`1s4^PYXATM

delta 454
zcmXBQze_>^6o%pB^=EoRt>A8Qms&y*1SxXRoPvr-Xee7GR~sCpC8emX;&O|ShH9v$
zhBT>YsG%Z|28SYoqNO7G4<fA>I`DA5;TsO8bg7hf6fF~;7#h+0Lt$ZjB6s(ZnP`3e
zNc-|yoTk;cV#%9^PpjVg`n<-gUeRrOdz5NG`S0gVf4oq9S4GZniZ!fa99?+yh%93X
z(@@cJi!^b8BC<%K)h6<QD^!riC_)IL<r2BUHFi-z(j`S1By_+JLBlC>f--W5pc@)m
z4v`l;;~0CGM;23<L>#^7#G75Dh6C)w#2jXk!<aqIg@hn}Z6fC|kbxIXtH=lHC?ks$
r5>WBUFKgJqA{G$C00N59h?$2mwfd9Tbem7xJ!X2##r9OchaJg3iH}i)

diff --git a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/data2vec_audio.py b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/data2vec_audio.py
index f3c7aa8a6..d669e1ed9 100644
--- a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/data2vec_audio.py
+++ b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/data2vec_audio.py
@@ -464,6 +464,8 @@ class Data2VecAudioModel(BaseFairseqModel):
                 features = self.feature_extractor(features)
 
         features = features.transpose(1, 2)
+        if prompt is not None:
+            features = torch.cat([features, prompt])
 
         features = self.layer_norm(features)