From 6ee505cffd36a7cbcbc97adb6cef62bfc671c3c6 Mon Sep 17 00:00:00 2001
From: mrq <mrq@ecker.tech>
Date: Wed, 12 Mar 2025 23:17:27 -0500
Subject: [PATCH] fixed dac

---
 data/qnt.dac               | Bin 0 -> 4328 bytes
 vall_e/config.py           |  15 ++++++++++++++-
 vall_e/emb/codecs/dac.py   |   2 +-
 vall_e/emb/qnt.py          |   2 +-
 vall_e/models/ar_nar_v2.py |   2 +-
 5 files changed, 17 insertions(+), 4 deletions(-)
 create mode 100644 data/qnt.dac

diff --git a/data/qnt.dac b/data/qnt.dac
new file mode 100644
index 0000000000000000000000000000000000000000..bb6a3e9d125c831e88e2a6486e258cad20d80a0e
GIT binary patch
literal 4328
zcmbtXcYMz0`o6FGiX!$FBcen?(-1{#rCM=HLXcA$Gu|XGLGmWw<c(1~p*FRv)G>~E
z?8YcnReRPrVnvl&RmV7LRE={>PyhIRet({PKOx`oJomh=`+8<XcWT$63uJ%{2}*Ql
zB%}v52?!e6J}4w0sCR0*H{IolOHEI7zk2)^S8|5io@XSv(%g0*TstISSit}G<2#ZG
zp3IcA!J!GM>F&^!%w%t0S9-c@uy0(tJ0aDR;Z4s>@cKIP$5+0d#6R9Th>|gmwnV(c
zv5rrSQy@4Bwg3B3qIYnb+t+z5Rz;-!J5Isz<HkCLVw}RkQH+H>8|wH*M@L7#w4eNc
zqTcm7MdHWzaEcCdiUr3Ouul?FEgMepBuUD*R3gdDl#FponHjqZN9`Nylur7?LS}uZ
z%qt6JW1Mn-T4?5!|GyT*>r{vz|NiY_BAc-pNxC01G*S;p7z-nhC#4F{VUNV|nl|S&
zZqgl^$np$hB3sH0Bx(eEYPjTRPrD9i9PY~kl-GB4KWCyNe;4HliPVRBUs5p~)iI79
z?nMTt>0tb(bFo_vz>ib30d8ZHZjoSJE{o-ihF}o7$|`-Nk;ug{?Wa5SZE1<dnxvZ%
zX}(_2i{|qMdwM}Hs7Ghy-)*zE4>grPB9zah4-@5kG_&7-;yFvoLTPKOk`!8>l^KUz
zeRcKJ9Nj2S@TCsrH{z@1wUn;adU6=&(EuYjQwHiy4dHPaiSzocd9%{uDMUAqT5NCY
zan9m#yVgk?KESJMg!*DIKE@GFw$%;o?+(%vPq>u*w6v~=pLyUW%-79mjhoU6t<X;!
z+wW!+VoRj)7H(iF?#fGj$S<|FTt_d-#ZUUWPC^2<AWWw4F~ay2Bd|hUavNLqu^ixN
z+~i}9lq?M9r<^Wf_$yZ+mXWfD???%~#!PLFBC?bfWQ=^P^97oQ5ojT9EyoY&hq>b6
z2jZ}kPDfX^L9+bF^|Hk}AVf>@sV?CqG{;$4pjj-Y#rR0(XaIcBRVT|`IiV3a$D$Hr
zyIj%VG1K<>5oI`6o-t0tm?T?JL0V~bi~l3JqQj*l7t2um#;M4Hzc$k%(oD;+zg0ib
ze44ADND1E2JGzPYnS#@@fNSJmcqZX;i|N>_et_NpZO?L+kB0`OQIciViq-0z!6~RE
z{dJpVKMQ-&896wH9nxEN;dKN_H-3f?+dEf!+CGDL4OKN-zCe3?fdaB!7g%JM(NWGy
zGp6Yf@koevm2~E!kM-^|tNucZxGp-&J_d6;AIf*6W#=@%(CadkJNXVDiyyZzkOlRs
zCZQtB%1ND~L!};TGFPf`s9a+;U4(3H!Or3<v}zaPMaySJMreW-M=-i+S61XI+~-{O
zKq9{3Qe7cCY@cOjHJtt&p`stL3oGdn*3%tQMfzJ$Rbi6mp_mM@8h4g%h?nJt@kI5N
zXt`+bE^#jA$>+R>2pra5P*!UrT75K6hw&)C;d5=Lk9blC%3=;OzmB7b`IralGIZ91
zR>cXNVL1MVH+h<uHILP$sYd8%WU{`jW>3kKC=F+6`8zxF7Go`Q3#`XCF<nP<n|8Bq
zx+h;5&3~e?JYw&B<b6FZd2D1T>ZbkqtL45Cwqu5M!*$IuHd-Q6Wt@D7bnB&1U8L2x
z8>87+o4{+adw54?$rXCFg5AHVV{nHfn5?ZO2nmMW-dJqC>?c1<FT;5}<5-OgjUjsB
zkX7pkIVZbx0z8(57rIFvV->Q@-@Tl!Ypuf1u+=iw8?{+X;&~cPv6eS*hT+y-*AS<<
z)<L82&@g$FS^7`?UQ@Uch6EXiPRPS5c_C5k$54Kvy<`d9(wu(kZ~QP;4@;ah)VJ`n
z5RRcGfDds|>u8XKS#0y!&$zOV?RwJsrH1?{xw71lQrcL#2A3j6Y8ZZfuu=zew0gB0
zS85AgfmUYkwnX3^Zo(D4rz1H@%7fTpi2P1pU1@q=3*ZOq^gB{Of7kcrICJEV{DLwF
zLtP!I4Ot!#xADzjeTh?+uVAT(r?{cx^<P?)`%xd$`L4EPfRw;^xydM<%tFjz7p-gA
zt0ew>i%sOERK!wQt=F`sz9DlBAtm*FrpRj+ds~^v1nq}8XfFPS(!KlvHSsa08C#4(
zF89b$W6xx(?NNCXv*=i#G(#&rq}j&Q-(w%P$TB%(2-_gzb)O7DDd}W=nZRtTRCVpn
zzcCF1S<kR=Sq5OMoM&xGKuKAvCpeU&&{x+;UCW@q&5V-hVYnY^48GcO2Q4TY5Jb10
zKpX4Bk66_BVW~J;UysQg$uo=L2(p;(%VZ6+JtrG>dgyRQq82~GJ=xAcsmf*Y9I9Ey
z14|eyTX`IHt(#n0N3tb`k2n;LR*=oyhuJ7#KCS05q)W7}Vz`u&+Geqwl#muS*Y4|g
zauN+Cm`9PSU$7}hGZ4P~tDL1v*5IgyTlbl-Xsb0P7!`3qTWe>Vo{jV}4#{O~)14@W
z%2ugEjF9rWOrIkgQD(cLjFbh^*80k&!TN8RA|nuJ6S|*O>I@vT(b+Q0Ix>yTn2CK7
zf@>ONtA5g#dPa_DSDk{NECa=Lri|tQ*~#ZN%M%Q}&2=DOa2S#dC0mSX)cW)-t-%Qx
zi=OfjyKou{tb4CWHJyz6nxge|x-Mh~{H#`W>lio7vA=XeX-vUJu9A)NncDP`&kT=E
z<liz?x5@?M_h3VBCGDj@7z3^nuNK3vHmzUN{W4P~$tA00SJ`D9*_h?Dsh(jOJ*D4j
zoEEbglAw#MIz@2Md^?M#(!{Wwhl_a1t%lP^Hh+h4wlQ=U<Gio%u3nH*#*Gbel0#&n
zby;`0E`jPV@97>?(mc!cB!(G(b+_EO@iuEA!dPoDic--PcMw5;{Eg%7%^aIo7bTFL
z(H0+Yo$>!JJkeP?NQKcFXqw=qG_b5c)-E>3vuwssml?+93%Og*n*Qi$T(Q<Xo2YdV
z!>i)M_hgEm(<-_fJ^3N)BOH|tAtSVd%+zEYGu)hDDVt)qP4#_^e)yE_WH?WuyUvHr
z2KJR<=1o<r&n=r;=X5sa+5BuTMYJxTOK<hsyj*Qu^15cDhCa|sTG*yTj!ZM9R9e25
zAVbbm-ZqR*l6qPST~JOV`JRrj*q#}Slzr7P)(=mZ#0Q4eH%tYb)gi1V8+D<6sVik0
zn;IXS<_qJSBsP*cn1B?0Y|*_TPo*L!vIy3j)m)jVzmRODcXbgqYP2cTVm8UdGN|Zn
zjNF>bEUq%>%Qcu~k!0E2C@xo|vT1;c+S-_PKjV4Urujd#9Sf-o$?9i{rzD1OIair~
zEiAWROApgUYmsBU+yMjSW8;>7Sk2<%V;TDx=lQnGlV7;ob`C&Asi4DkrrrH&96DHi
za8D<3jqWo3XsVxMxKuSw+=X>*inNzfrnj#1nCa1MoBnTdouS_cdvvA_!RJ=}$Hr}+
z$UrJL7&=>FH?H!8)YZz2MxLqQ=B5&>3S<=PsK5C)z~=CJ({nD<x*s6dG)kb{HKu-I
z3aO2h<st;424U)?JKIa5^<=WCt4MsU**q&<WrZ=pd}(GYHk-1{<5q6e;%p$4*Q6-N
za2!rbdt9}NbD7<YXLoRtcCqQ&(s*qKpXmfs|4ng6zvDTZ@*8Z9tu%HoKwr}^Q@H@+
zaf561i1ffI`AyxrNxwq4Ovhq=Yn5H9A2Qq6V420x-C{V1IN4<S;)*GQI;>(?s>27S
z)&r!FUczF1Auh{b3<_&g)-c5~NE)kYarNgVLs$>2GW~Q5t>vmTmz!FJRiUP=$|67t
z8~zVSZ#>iPs3&`E8Xl6O$mVd&x7Cm7$0KsabWC@i#2OsH1UX?mxx_f7I%>*jW4mE8
zn?CAd0pqS5Eo=&TH1F}07Bt277g@(Trq9b6B620$_;&}VFhi%xayDlW{A8}Uao6<J
zbF;9--afGN)lTcQ?KYDyqKaOJ+cbOuw#9szp!1M{cx=%1GS3)kr8Z)DsmXs>M!V=K
zNiofsDLLG%>#W1anZj6an(2jJv{_xt)b%4%2_5NF^g4d=v4!kJnd0`k5?x-GQ)!q}
zIXJe2JxWdQ+o!L`l^mDs_Vn>4IaMOpO@zc2u~!+cl(b}bT)NBacB;0ESe?H`Ur$=5
zH!d;W@&EH&TJ6=jG$6*Qo_{VK<GlXgGimPzPK{S*(wZ?&pq)vR{{LyTR{Z#uKGPa^
z-{k~(o#6P`qGmK9Dbv&cPiJaJ?aMcskmT}s+{qbEohZZ>w6D@!iHUtZeVjMO<$Idw
zN{AcaPS5C@>TyEy&(2|?^+M}8Z^ruA$6og!uM-+u+5T+rNpJ_a`ndXfLIT`_65Poa
zwKrg3YI^@rr*6Kfw4_vzJH^i3^<wMTf`Ny}9h&6H2srrZ!9|D19h_(DUj5fU<KW!n
rgL7Z4wcGiJKehE?`RkKio<5nbK5nOe{?Xs<ao)<zba|Zy@uB|#L6vJ?

literal 0
HcmV?d00001

diff --git a/vall_e/config.py b/vall_e/config.py
index 363e003..e9a1dfa 100755
--- a/vall_e/config.py
+++ b/vall_e/config.py
@@ -318,7 +318,6 @@ class Model:
 	name: str = "ar+nar" # vanity name for the model
 	version: int = 5 # 1 = old with MultiEmbedding, 2 = new with AudioEmbedding, 3+ = additional embeddings
 	size: str | dict = "full" # preset string or explicitly defined dimensionality
-	resp_levels: int = 8 # RVQ-bin levels this model supports
 	tasks: int = 8 # ["tts", "ns", "sr", "tse", "cse", "nse"] and leaves two more for anything else I want (like "svc") (unused)
 	langs: int = 1 # defined languages (semi-unused)
 	tones: int = 1 # defined tones (unsued)
@@ -382,6 +381,16 @@ class Model:
 	def tokens(self):
 		return self.audio_tokens
 
+	@property
+	def resp_levels(self):
+		if isinstance(self.size, dict) and "resp_levels" in self.size:
+			return self.size['resp_levels']
+		
+		if cfg.audio_backend == "dac":
+			return 9
+
+		return 8
+
 	@property
 	def audio_tokens(self):
 		if isinstance(self.size, dict) and "audio_tokens" in self.size:
@@ -1043,6 +1052,10 @@ class Config(BaseConfig):
 			if not isinstance( model, dict ):
 				continue
 
+			# was made an inherent property tied to audio_backend
+			if "resp_levels" in model:
+				del model["resp_levels"]
+
 			# to-do: prune unused keys in here too automatically
 			if "experimental" not in model or not model["experimental"]:
 				model["experimental"] = {}
diff --git a/vall_e/emb/codecs/dac.py b/vall_e/emb/codecs/dac.py
index ff6f660..0eddcd2 100644
--- a/vall_e/emb/codecs/dac.py
+++ b/vall_e/emb/codecs/dac.py
@@ -2,7 +2,7 @@ import torch
 
 from dac import DACFile
 from audiotools import AudioSignal
-from dac.utils import load_model as __load_dac_model
+from dac.utils import load_model as load_dac_model
 
 from typing import Union
 from pathlib import Path
diff --git a/vall_e/emb/qnt.py b/vall_e/emb/qnt.py
index 458d332..7242a8c 100755
--- a/vall_e/emb/qnt.py
+++ b/vall_e/emb/qnt.py
@@ -117,7 +117,7 @@ def _load_dac_model(device="cuda", dtype=None):
 	else:
 		raise Exception(f'unsupported sample rate: {cfg.sample_rate}')
 
-	model = __load_dac_model(**kwargs)
+	model = load_dac_model(**kwargs)
 	model = model.to(device)
 	model = model.eval()
 
diff --git a/vall_e/models/ar_nar_v2.py b/vall_e/models/ar_nar_v2.py
index a93e3c1..369ee1f 100644
--- a/vall_e/models/ar_nar_v2.py
+++ b/vall_e/models/ar_nar_v2.py
@@ -828,7 +828,7 @@ def example_usage():
 		available_tasks = ["tts-nar"]
 
 	model = AR_NAR_V2(**kwargs).to(cfg.device)
-	steps = 500 // batch_size
+	steps = 250 # // batch_size
 
 	optimizer = cfg.hyperparameters.optimizer.lower() if cfg.yaml_path is not None else "prodigy"
 	scheduler = cfg.hyperparameters.scheduler.lower() if cfg.yaml_path is not None else ""