From a0fb9639b0354d85124b150254f7880437be53d1 Mon Sep 17 00:00:00 2001
From: dohe0342 <kimdohe1070@gmail.com>
Date: Thu, 26 Jan 2023 14:19:49 +0900
Subject: [PATCH] from local

---
 .../.train_adapter.py.swp                     | Bin 81920 -> 81920 bytes
 .../train_adapter.py                          |   9 +++++++++
 2 files changed, 9 insertions(+)

diff --git a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.train_adapter.py.swp b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.train_adapter.py.swp
index 2d201c0c33c74ca78b2e7b8e32e7ebd2c27047bf..5f6b4aa6203ad6d4d0cdbad518240ee456b5e19d 100644
GIT binary patch
delta 418
zcmZo@U~On%6-_b-^Ym4))H7fJ0s#hwe>|6xKWr4;Ex?$w`J#Y4Bcs^nkAlK_yfr)w
z3?G0Pq(*GApo6-Z7#{-z7Z88oWnef1#EXG=5fD!S;!Ys$0OD96HUVN|ApQzAVRN2C
z6VK#pzH{qSD>6$oG#M0tpr9x-uSBCDN0R}}1q<awm!wvdL>t7$D%dKhE2!&b<!9z;
zfW-{0Vj=oK3PB=36(CBlxTGkvKtofnxF82^c3OTBL``O%LP1VENL{fN&>DH5wK+hI
z)wMtlhz*efisToh=4q%;ELE2&Nsdp-OxG*OQP)vWFH+akD@sjF$;r%1E!NQ7oaZ;q
tk#`#~M4Ff&!IHCCaKd&*M$O45Cq^iN%!C035CfY961Xhb+%&0fEdcjzaf1K=

delta 234
zcmZo@U~On%6-_b-^Ym4))H7fJ0s#hwUp$wRb2p0a7GTWYd{IE2k&$onM?qmd-X>lK
zhHF3!Qo}b{&_UfSkB@=D8HlBTSQ3bDLsc9B((8a2#9s@<#Xy`3#5q8$#k)Drp^0bm
z*>5V7E#qP*t9`fBQGkMiqRc!n6T~RdP}fxl@!%XSV*^cv$sc^?PTu~5cXNYZizDxT
lpmW-pAa2UuEI461BctWylM^GRPq1f<WdZ5l+%&0fEdW5;JW2on

diff --git a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/train_adapter.py b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/train_adapter.py
index 32fad2147..4901d5fb7 100755
--- a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/train_adapter.py
+++ b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/train_adapter.py
@@ -834,6 +834,7 @@ def compute_loss(
 
     texts = batch["supervisions"]["text"]
     print(texts)
+    
     #texts = batch["greedy pseudo text"]
     token_ids = sp.encode(texts, out_type=int)
     y = k2.RaggedTensor(token_ids).to(device)
@@ -1582,6 +1583,14 @@ def run_adapter(rank, world_size, args, wb=None):
 
     #train_cuts = librispeech.train_clean_10_cuts(option='male')
     train_cuts = librispeech.test_clean_user(option='big')
+    pl_texts = open('/workspace/icefall/egs/librispeech/ASR/tc_big.pl', 'r').readlines()
+    pl = {}
+    for text in pl_texts:
+        text = text.strip().split()
+        pl[text[0]] = ' '.join(text[1:])
+
+    print(pl)
+    exit()
 
     def remove_short_and_long_utt(c: Cut):
         return 1.0 <= c.duration <= 20.0