From c40cb1a1a1102d568ab73f131e0eb765126733a2 Mon Sep 17 00:00:00 2001
From: dohe0342 <kimdohe1070@gmail.com>
Date: Tue, 11 Apr 2023 15:32:55 +0900
Subject: [PATCH] from local

---
 .../.data2vec_audio.py.swp                    | Bin 40960 -> 40960 bytes
 .../.prompt_tuning.py.swp                     | Bin 81920 -> 81920 bytes
 .../data2vec_audio.py                         |   1 +
 3 files changed, 1 insertion(+)

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 903071da0f289d19507f5b0396cbb8e3b30a2391..288955dc0594d7e01703967e63272e726cc355d8 100644
GIT binary patch
delta 174
zcmZoTz|?SnX~P5oM(52F1&Z}~xmg()jM*3%7z7v?oF@x9Jl1RFU|<LU;+O0U4D*3_
z9uOA-aRCtf05Jy;9|Y>C0%Bvf%}S1vtedC1g)#XS6y@g@lqlFLKp1+d6$OcTDH<AS
hsfi_}MXAMl#hF#98k*4tu{tm%I7JLM-}6uk2LLZ;FI@lt

delta 104
zcmZoTz|?SnX~P5oMyJga1&Z}~vsoD!`q>y57z7v?oF)r8Jl0#o!NA}O#J|}Y7=8iq
wSs*?G#4CU}1&C#V_!>|LNLLH{W+g{S*2y2{hD<)_Cc8Pn&5>#ITMyxI06YyEi2wiq

diff --git a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.prompt_tuning.py.swp b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.prompt_tuning.py.swp
index 8c057dcbeddd2123893db9fcb18e9509690a2135..a0cce3133e934abc3fd3ca5e53b1d64f7bba9b55 100644
GIT binary patch
delta 35
pcmZo@U~On%6-_b-^Ym4))H7fJ0s#hwy1ynVZyQCoiZZ^o2LPmv3Ge^_

delta 35
pcmZo@U~On%6-_b-^Ym4))H7fJ0s#hwioYf)OBzMDiZZ^o2LPe{384T0

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 cbfc809e9..d2513b298 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
@@ -491,6 +491,7 @@ class Data2VecAudioModel(BaseFairseqModel):
         ## for prompt tuning
         if prompt is not None:
             #features = torch.cat([features, prompt])
+            prompt = prompt.expand((features.size()[0], prompt.size()[0], prompt.size()[1])
             features = torch.cat([prompt, features])
             prompt_padding_mask = torch.zeros(promt.size()).type(torch.BoolTensor).to(features.device)
             padding_mask = torch.cat([prompt_padding_mask, padding_mask])