From ff1ad015005157e97df223685e82ce19bcc480a1 Mon Sep 17 00:00:00 2001
From: dohe0342 <kimdohe1070@gmail.com>
Date: Thu, 6 Apr 2023 22:21:34 +0900
Subject: [PATCH] from local

---
 .../.data2vec_audio.py.swp                    | Bin 4096 -> 16384 bytes
 .../.data2vec_encoder.py.swp                  | Bin 16384 -> 20480 bytes
 .../data2vec_encoder.py                       |   6 +++++-
 3 files changed, 5 insertions(+), 1 deletion(-)

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 7341b56deb310af548f56588fb65a84f3def7ba9..2ac3e2edb86139a6b0d198d42404b72de079869c 100644
GIT binary patch
literal 16384
zcmeHOUx*|}8Sm(wi6*CsBFTdfb*_+}bGB!8E;(??oZM}0WpjU%Tq1IuO>cM2Y~S^C
z_f~c9?yM4Dd<ll!;e#*21w{lwNk|ZtNC@a#5Q(7p2Sv~m1pfsLiW=kZtE&Dtv$^v<
z-SC^~sjup*uYUFQ*Ii%rtoI*ZJIy}k^er5BTGo-j9|}JI^(*_ne8(+TkZkb&;rj+8
zy3z3<>hU~WJni|%ljQO_kvLJ}d(mv0CklnV?L=Ibsps=v=<|^mMLj;2y(nB4p-egV
zH+sh&eWaI)Jm!I$2``pG?sMVF%*%MhrF4VAmb*3R1zzS2wz%(lc@QQ}I-Tva{oOLq
zGEgvZz&bowQkVL|hYzw3y#KL6(0;cJv<$Qiv<$Qiv<$Qiv<$Qiv<$or7|7CFt<Pbc
z?=WNCG2ge$e7|W9Y08U*c2V!YM$V1=Pc!Y`F!GTpf7moE+W*JMg^_<}M*eRj9~*hc
z7+kb}rz!uC`Mx?MH^uE&%RtLO%RtLO%RtLO%RtLO%RtLO%RtLO%fQ=`0WYvDS~K56
z>t*c!7wi8Yz0a~<0iFcT0%w4uKo3{|{`FqV`VH`F;D^8ufaic`fv*Bj0TFN#cnG))
z_~`-E0j>g{1p;6lcn}x>3&26(jXNysSHLfU=YVH{&j7o?3E*zvpSN4q_kkCIr-2MO
z4E*Iimi1@gYXAq10sDbh-)&hh0M7$Y02hGAfct^_fCIprw^`O}z|Vk}fNucL0M~#i
za1wY3xCi*#t(Nr*;OD^ez!l(Wzy*!~|9F>W{So*!5C8+<Bf#tXVH@y8;9=k#&;#xQ
z?gU=nXIZ}oeh0h&c)$mN*O1`&3UCGZG;k4E2KEE{fa{pQ9|PY5z6)Ffo&vrATmnu3
zCxOEN&Gp^DjSgmxaKutCa3da%vki8X^_Q(ew7~K>NJtz6VLY}<!44a;9Y?*kktUMc
z{VrSTl+p!up6#&XzzSPDs!ph2rD{(dkPcC{ffa&<lEq2J&LlBkZdRrS>w<fi%LY_J
zhagidb!tAKDrCbU%f^`-c~dU-&_z|1%oSYbQ6?QPO?e#H_D+}Wc3C7k)p!c!O7YPl
zn}qRODyC6M9*rC?@Y0M6NeXpOVK5BmmClz?cW6XfTrbOFm-q+$?!=RqhxH#_7IVBl
zv(3ZtMka@=UL<+fD$Fguq081of^EtayHN+jBG#4Fn9?vtUn~*K4n&eBdFG}fS@+h%
zD9ont4^>a8A$h0r7V2-WCPNmd4&Lwyl8O1UI}DvR6_MvLjsdWx(z+3>2n~HKj2Vyf
z2^V-l$gbK-SWq?~XCUt>3mS6_;f0G;<(oM%Ik56ua^ljXo9$N}Gh&+Nbg0ZSC0$mK
zsf&wW%?a;-G*+5RKj>&qveGK}W*!R6w&a_1vAPquZJ0*9%Ytw+TwLn1EicOX@Vv;Y
z@y{!B!2;B~SyF1K+=2LSPH5ed3(H-$NNshjSiT5V)LYa3dJ9z8abmvh?sAbxTwIO}
zcM&c+U6iZSr4G)6Y?|_6mI!~tSx=Ivq6ltMb*Nkfe$=3cf`$$liaC7DQ);dz=VlxZ
z3J&{Lp^T=r9Py2)TcCO@OIWtS;o6%ykA2QeRK*POScXYlc@K3^1`F+`aUv%6PGj)p
zg;~{bN1>2e&FPgNvyZYJ*H7ZD%5wFSW<Z_PqBUG)Q1zn4sHfZv;apY6R8`?!HR#OS
z9>xLRann3g-d~GSUDmHU&~sQuusx!DXBlLahL_=W*sp?F8O(<?*;RSn+h#<Vx|&uR
zoo*KJ5#u`;ou9cQ?qxY{0q%B~ZMY_8+to+~hd8Ei+=wLF*3x41^MA~x;&*l)8Tqp+
ztQ@(amvq;Uir0E+{k)>8JhSTZz0Mmn)_-)XdD*KSm4O_qkj)ZRxYbSgrsgy)2>9Ec
z2vm4Sh)~yF={S0NHCg7o9H9DLzY{E^4!t>w{^Qs==Lv7ZaW}SgjG>!RY^OQ`bvjO#
z*!TK*;N9El4f;;S@C}vdkhYhSN&$__7}6WErM_cP2_1B6OX;4byrOtmaMHq3Y1c>7
zsDv>+!etobUUW=9WSmZdJmNO))0bSj8Qi~AyS`<fV&EOeb#Z6UT-Wx*SZWiA^0HW6
zAdG~|?33Qs;BoE`R+4xWj$w%(c~T<S<2_wij+^j_*iQm3i~;t%gv^hU@tD&+nNILj
z;f#59N|nJLiF3ZiBfAi;omoALATtulgi$0*qC7*$WZ@)DM8;rCQKZLfs-jXPi%M|a
z`khTH8+nMBeA8*vEMu3Yxj02nif1r39V9*>yx_FH!%n+RH50{l*`;k1%u}E>HM!Jf
zG?rBzdQRv_gn?pIJeuzd_he@q^MQAui&5Hh2BA_RJnl)psz<CfGQr28g!j9!oJY;J
zbdjhl#ci<TCliEqt{yqzHvwMa#=fwb34IX-&?&-o)C!m<@em@jW(8b$qdfL$4!ww~
ztj@}*Fpe9gWhZ#lOQ)(rY9>0F@UaKSz~jC4h_y8%!Q%rxvzp?wsqRt_lm@80mzNBc
zMO`)uc@z`|OuTHPI5$pN{KRE}OH*<zVcDs*l@n(kJ%P3nrz@Cl_#(j*bb#kGALL9~
zGD0aHOk|P^JOyNea|bQY*UISfp^zVe+E{`^EVXd0kh0Tj=Zj**36z^&tOe*;Nz$pH
z=Y3}T9kz-$?D52Oxy#mK-%&S{Fq6z1!MG3)%F?lD{eKYa>2CqF{x9~=e~UH#%fK_h
zHNXWNfB}ENTK+ol3UC#85|{v=0zMAdz-_=^v6jCM`~-Lv_!6)UxWEIz?ZAts?{8B7
z?a?yOGSD*6GSD*6GSD*6GSD*6GSD(mFrbqo`m>ovp|VJ&O=w48kKF*|TPY*qsAQT{
zI}9OvfOb5j*$1+;Uj3<3D$O=JS%gJ1k}0RKoUQWWNMTlLnhDH8r`9*2j8kI=LY0p~
zEP-hs<x*Wt=@-E#$rgu*BO$4_UTBpL1EfiURsGK8xJ+^q>C19tP0kb*L-nci4u!F0
z+B1oxsoF7U%GFL2eKd<3iNUf@QmPlaRc+XWk?OIVoSv$WtPoYFgM`@SQY++UqOZ$T
z|9U&CmhZEu*<RV(TEdp97+!RmLT-x8o;q8euqidx?;umY#21froC)`0yCQ6U?_*vM
zw7gYw>X^hswu?+z4_j_NTg0v?KI~n|4jn?y-4q`xg)=#I%Aq*;uRJy->EG~ZXA*U!
PoR0YlmHm6l{f_lNM*hsW

delta 7
OcmZo@U~EuWAOHXgJOWYx

diff --git a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.data2vec_encoder.py.swp b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.data2vec_encoder.py.swp
index e83ed7901eb89b97ed90525fed071d34d59da0d6..62735843d34cf79d9d6656a30d12711c6b24f3b5 100644
GIT binary patch
delta 692
zcmaLUPe_w-9LMqR<DM;-(`|YfOVVExa;ItdhoroQ1{G9LAh@2~(|l-qp85D}&}k#W
zphE{oi3k!z(H{um!7g>{5ELDvQ-Oy(b+bBTc=A169{a(I-@niA$L~+Arqv#4Yg++L
z{GWl&NHic-5otlAOw{ACL|otR$Uw86@RPCk(M6@V7egY?4>24g>&I1>Rvm8@>A_!y
zKcg|diWkV@6he5z(tEHlg>y)wA3uU3AMpXta2E?GBaibKL=sz^?j6>!tjh%{(kNgW
zLx^HCz+vzhOSpvrWU$R;|6m2vIE~+0^Qz(}QZIv5rBu<FznPtxHEh~@rRe0$LdKZ!
ztcu}Tj-Ac9PRS{IYFrIwPAF-{)PC+@xosA*1=G%Zvjj(n{}(H{&h@?Bw2}H)`slII
zk%M$!pDMX##k4);R804V<$0E!_ovcN5>+E75)6xMpz$&PzHK`y@)Zy90B3O=zdA&o
zqKwNpgOlh&C%&|cyv91}a50B#xQH=yL&vv>T#{E2UK4k46<08c9_+M<e8OAY$8F?r
d1W9Zj7Ws|^OrjqfVPDhR!i%X^zc<w#+6DAZd?^3`

delta 305
zcmWm9Jxjw-6vpv$oAd?jOKOCGb-2($OF<)a5_EB~ppzB^7xfD$?FVR}9Ysj8IY^f-
zR#3Q$gNUPxsDq=RgW%RouKLe`9~_=%IVTvYAW*kcOLa_Q%@!*e>5E7P>Ga~N<C!hz
zt`a6X(spTu&`Vo~(p0sp`fCM|Rk(P~i@acnLp1T36S=`T0(7y1ZIqB?MW%ShH72;g
z8G7ho19SMXMc(m%2%#;TGGN(93wiunB2T!-CHAn31=R4!{R12!{pC-EU-R)tyAWSB
hicxdMe79@4_-)ZGpLS1<^}6<#eO<3>bFeh_{sH+iHbejb

diff --git a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/data2vec_encoder.py b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/data2vec_encoder.py
index 819667c41..b3c482c1d 100644
--- a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/data2vec_encoder.py
+++ b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/data2vec_encoder.py
@@ -65,7 +65,11 @@ class FairSeqData2VecEncoder(EncoderInterface):
         )
         model = models[0]
         model.feature_grad_mult = 0.0 ## for conv network freeze
-        model.mask_prob = 0.0 ## for conv network freeze
+        ## prevent overfitting
+        model.mask_prob = 0.65
+        model.mask_channel_prob = 0.5
+        model.mask_channel_length = 64
+        model.activation_dropout = 0.1 
         
         self.encoders = model
         self.pretrained_params = copy.deepcopy(model.state_dict())