From b70122039d0c5f7e1589ba2aa66f9cfcb684524e Mon Sep 17 00:00:00 2001
From: dohe0342 <kimdohe1070@gmail.com>
Date: Thu, 2 Feb 2023 19:18:48 +0900
Subject: [PATCH] from local

---
 .../ASR/transformer_ctc/.decode.py.swp        | Bin 53248 -> 53248 bytes
 egs/aishell/ASR/transformer_ctc/decode.py     |  21 +++++++++++++++---
 2 files changed, 18 insertions(+), 3 deletions(-)

diff --git a/egs/aishell/ASR/transformer_ctc/.decode.py.swp b/egs/aishell/ASR/transformer_ctc/.decode.py.swp
index e80ac848bd3ab0bccf7cc48f9c4f2e225c9cea7b..d53020f6520d845e7f3574f45358a684df9135e9 100644
GIT binary patch
delta 646
zcmY+<%}N4M6bJC*w-tdwxHB#oIObqRi=vPaY#FUwiI~|njZtJA``S30XcbWs6m4u1
z21UX222q<96}0XNgdQM>{&#c?2Yz$UoOAD;bM%%&Z#f>`eYTYG;hQ}k^|^?2Hxc=W
zQq8f}W#eXq=o}7V2MWMp0qpQOOmqSVkcA{fzz;6y3=zG-BV52XD6j_e5CA89dx^f_
z8QO3Js+X2%6~QtjApk!fq9?co9TG4J9_XO;S9pOl(A~z?mEYB1c&citqDgtKvMDwb
zi-k!f#b=yJwfcH3n^Tm?23NX4sba2SHqG+IoSLoF%jN;|B~FURs<3#RDZC=7ywJV2
zTV||xOM^w?A|@*@R_j-yCgOpj)sQHPc%Ya@*Rq+G&6BdYJA*mOTvJQ=T6B|ZS?rOD
zdsUgCsl~yrkilk+?`pMlDHBzBRmtVK6k;Kng+haSgQso3E!gz8=WIb^rJZpk%~t*=
iXE8I;n-uqUzx&f!iT*?)VchK}oXv^rapU*i&i(*S^OT$b

delta 286
zcmYk$F-QV&07mi8F7rAOE)XGv5FE5ABB7<I79(+Rj}{lHAtY4L2wg3x;oMv<C~L~(
zKQy#7H<^Q@TW#Ue(h@zIdhpv{%ePv-^*D3`S@Y(uT}58zq_2=9SIT+|ez(~SOS@Ra
z*pXgwgEq?8M-Hoq;(K12;28tl;tWOXA%i3weA?0oPw3zrZ4|Ks$JW;0PL-t!HnD*?
zK3ILnB~H;m9VO(E!8B-&u46%SQL6{M!b&JsJ~=$Dq*JM7CI00N@79fde_)x1#`dhY
L)Qg&*{!HQ*J(4$}

diff --git a/egs/aishell/ASR/transformer_ctc/decode.py b/egs/aishell/ASR/transformer_ctc/decode.py
index 4fc8ecb26..f69128454 100755
--- a/egs/aishell/ASR/transformer_ctc/decode.py
+++ b/egs/aishell/ASR/transformer_ctc/decode.py
@@ -453,6 +453,12 @@ def decode_dataset(
         num_batches = "?"
 
     results = defaultdict(list)
+
+    subs_all = 0
+    dels_all = 0
+    ins_all = 0
+    char_num = 0
+
     for batch_idx, batch in enumerate(dl):
         #logging.info(f"decoding {batch_idx} th batch")
         texts = batch["supervisions"]["text"]
@@ -483,9 +489,18 @@ def decode_dataset(
         )
 
         for i, hyp in enumerate(hyps):
-            print('hyp = ', hyp)
-            print('ref = ', texts[i].replace(' ', ''))
-            print('')
+            #print('hyp = ', hyp)
+            #print('ref = ', texts[i].replace(' ', ''))
+            #print('')
+            ref = texts[i].replace(' ', '')
+            [cer, subs, dels, ins] = metrics.get_cer(hyp, ref)
+            subs_all += subs
+            dels_all += dels
+            ins_all += ins
+            char_num == len(ref)
+    
+    cer = (subs_all+dels_all+ins_all) / char_num
+    print(cer * 100)
         
         '''
         for lm_scale, hyps in hyps_dict.items():