upload files
Browse filesThis view is limited to 50 files because it contains too many changes.
See raw diff
- README.md +1 -0
- data/lang_bpe_500/HLG.pt +3 -0
- data/lang_bpe_500/L.pt +3 -0
- data/lang_bpe_500/LG.pt +3 -0
- data/lang_bpe_500/Linv.pt +3 -0
- data/lang_bpe_500/bpe.model +3 -0
- data/lang_bpe_500/tokens.txt +502 -0
- data/lang_bpe_500/words.txt +0 -0
- data/lm/G_4_gram.pt +3 -0
- decoding_results/1best/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/1best/errs-test-other-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/1best/log-decode-epoch-40-avg-16-use-averaged-model-2023-06-13-19-45-46 +26 -0
- decoding_results/1best/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/1best/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/1best/wer-summary-test-clean-epoch-40-avg-16-use-averaged-model.txt +2 -0
- decoding_results/1best/wer-summary-test-other-epoch-40-avg-16-use-averaged-model.txt +2 -0
- decoding_results/ctc-decoding/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/ctc-decoding/errs-test-other-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/ctc-decoding/log-decode-epoch-40-avg-16-use-averaged-model-2023-06-13-19-42-20 +31 -0
- decoding_results/ctc-decoding/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/ctc-decoding/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/ctc-decoding/wer-summary-test-clean-epoch-40-avg-16-use-averaged-model.txt +2 -0
- decoding_results/ctc-decoding/wer-summary-test-other-epoch-40-avg-16-use-averaged-model.txt +2 -0
- decoding_results/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/nbest-rescoring/log-decode-epoch-40-avg-16-use-averaged-model-2023-06-13-17-24-42 +6 -0
- decoding_results/nbest-rescoring/log-decode-epoch-40-avg-16-use-averaged-model-2023-06-13-17-28-30 +180 -0
- decoding_results/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/nbest-rescoring/wer-summary-test-clean-epoch-40-avg-16-use-averaged-model.txt +21 -0
- decoding_results/nbest-rescoring/wer-summary-test-other-epoch-40-avg-16-use-averaged-model.txt +21 -0
- decoding_results/nbest/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/nbest/errs-test-other-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/nbest/log-decode-epoch-40-avg-16-use-averaged-model-2023-06-13-17-21-29 +35 -0
- decoding_results/nbest/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/nbest/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/nbest/wer-summary-test-clean-epoch-40-avg-16-use-averaged-model.txt +2 -0
- decoding_results/nbest/wer-summary-test-other-epoch-40-avg-16-use-averaged-model.txt +2 -0
- decoding_results/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/whole-lattice-rescoring/log-decode-epoch-40-avg-16-use-averaged-model-2023-06-13-17-33-26 +251 -0
- decoding_results/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt +0 -0
- decoding_results/whole-lattice-rescoring/wer-summary-test-clean-epoch-40-avg-16-use-averaged-model.txt +21 -0
- decoding_results/whole-lattice-rescoring/wer-summary-test-other-epoch-40-avg-16-use-averaged-model.txt +21 -0
- exp/decode.sh +16 -0
- exp/epoch-40.pt +3 -0
- exp/export.sh +10 -0
- exp/jit_script.pt +3 -0
- exp/log/log-train-2023-06-01-20-30-01-0 +0 -0
README.md
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
See https://github.com/k2-fsa/icefall/pull/1111
|
data/lang_bpe_500/HLG.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b5dbbe8b485c0cb37d11e07e8e734990f1e40a2d00fe9689d8da2e7b6fe72883
|
3 |
+
size 845007583
|
data/lang_bpe_500/L.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b1b88996f918737fba67fbd29152018b51a537c16ce0718a2b43d5140583224e
|
3 |
+
size 19025703
|
data/lang_bpe_500/LG.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3bb9f021c7aad79d45dc275ba8154a430c4f660a319dcb872cd52500f25553d6
|
3 |
+
size 249852195
|
data/lang_bpe_500/Linv.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:cbc8b3687a1b8f0811a84106b3b310642566c7b1bc282a929878f9269507a2c6
|
3 |
+
size 19025703
|
data/lang_bpe_500/bpe.model
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c53433de083c4a6ad12d034550ef22de68cec62c4f58932a7b6b8b2f1e743fa5
|
3 |
+
size 244865
|
data/lang_bpe_500/tokens.txt
ADDED
@@ -0,0 +1,502 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
<blk> 0
|
2 |
+
<sos/eos> 1
|
3 |
+
<unk> 2
|
4 |
+
S 3
|
5 |
+
▁THE 4
|
6 |
+
▁A 5
|
7 |
+
T 6
|
8 |
+
▁AND 7
|
9 |
+
ED 8
|
10 |
+
▁OF 9
|
11 |
+
▁TO 10
|
12 |
+
E 11
|
13 |
+
D 12
|
14 |
+
N 13
|
15 |
+
ING 14
|
16 |
+
▁IN 15
|
17 |
+
Y 16
|
18 |
+
M 17
|
19 |
+
C 18
|
20 |
+
▁I 19
|
21 |
+
A 20
|
22 |
+
P 21
|
23 |
+
▁HE 22
|
24 |
+
R 23
|
25 |
+
O 24
|
26 |
+
L 25
|
27 |
+
RE 26
|
28 |
+
I 27
|
29 |
+
U 28
|
30 |
+
ER 29
|
31 |
+
▁IT 30
|
32 |
+
LY 31
|
33 |
+
▁THAT 32
|
34 |
+
▁WAS 33
|
35 |
+
▁ 34
|
36 |
+
▁S 35
|
37 |
+
AR 36
|
38 |
+
▁BE 37
|
39 |
+
F 38
|
40 |
+
▁C 39
|
41 |
+
IN 40
|
42 |
+
B 41
|
43 |
+
▁FOR 42
|
44 |
+
OR 43
|
45 |
+
LE 44
|
46 |
+
' 45
|
47 |
+
▁HIS 46
|
48 |
+
▁YOU 47
|
49 |
+
AL 48
|
50 |
+
▁RE 49
|
51 |
+
V 50
|
52 |
+
▁B 51
|
53 |
+
G 52
|
54 |
+
RI 53
|
55 |
+
▁E 54
|
56 |
+
▁WITH 55
|
57 |
+
▁T 56
|
58 |
+
▁AS 57
|
59 |
+
LL 58
|
60 |
+
▁P 59
|
61 |
+
▁HER 60
|
62 |
+
ST 61
|
63 |
+
▁HAD 62
|
64 |
+
▁SO 63
|
65 |
+
▁F 64
|
66 |
+
W 65
|
67 |
+
CE 66
|
68 |
+
▁IS 67
|
69 |
+
ND 68
|
70 |
+
▁NOT 69
|
71 |
+
TH 70
|
72 |
+
▁BUT 71
|
73 |
+
EN 72
|
74 |
+
▁SHE 73
|
75 |
+
▁ON 74
|
76 |
+
VE 75
|
77 |
+
ON 76
|
78 |
+
SE 77
|
79 |
+
▁DE 78
|
80 |
+
UR 79
|
81 |
+
▁G 80
|
82 |
+
CH 81
|
83 |
+
K 82
|
84 |
+
TER 83
|
85 |
+
▁AT 84
|
86 |
+
IT 85
|
87 |
+
▁ME 86
|
88 |
+
RO 87
|
89 |
+
NE 88
|
90 |
+
RA 89
|
91 |
+
ES 90
|
92 |
+
IL 91
|
93 |
+
NG 92
|
94 |
+
IC 93
|
95 |
+
▁NO 94
|
96 |
+
▁HIM 95
|
97 |
+
ENT 96
|
98 |
+
IR 97
|
99 |
+
▁WE 98
|
100 |
+
H 99
|
101 |
+
▁DO 100
|
102 |
+
▁ALL 101
|
103 |
+
▁HAVE 102
|
104 |
+
LO 103
|
105 |
+
▁BY 104
|
106 |
+
▁MY 105
|
107 |
+
▁MO 106
|
108 |
+
▁THIS 107
|
109 |
+
LA 108
|
110 |
+
▁ST 109
|
111 |
+
▁WHICH 110
|
112 |
+
▁CON 111
|
113 |
+
▁THEY 112
|
114 |
+
CK 113
|
115 |
+
TE 114
|
116 |
+
▁SAID 115
|
117 |
+
▁FROM 116
|
118 |
+
▁GO 117
|
119 |
+
▁WHO 118
|
120 |
+
▁TH 119
|
121 |
+
▁OR 120
|
122 |
+
▁D 121
|
123 |
+
▁W 122
|
124 |
+
VER 123
|
125 |
+
LI 124
|
126 |
+
▁SE 125
|
127 |
+
▁ONE 126
|
128 |
+
▁CA 127
|
129 |
+
▁AN 128
|
130 |
+
▁LA 129
|
131 |
+
▁WERE 130
|
132 |
+
EL 131
|
133 |
+
▁HA 132
|
134 |
+
▁MAN 133
|
135 |
+
▁FA 134
|
136 |
+
▁EX 135
|
137 |
+
AD 136
|
138 |
+
▁SU 137
|
139 |
+
RY 138
|
140 |
+
▁MI 139
|
141 |
+
AT 140
|
142 |
+
▁BO 141
|
143 |
+
▁WHEN 142
|
144 |
+
AN 143
|
145 |
+
THER 144
|
146 |
+
PP 145
|
147 |
+
ATION 146
|
148 |
+
▁FI 147
|
149 |
+
▁WOULD 148
|
150 |
+
▁PRO 149
|
151 |
+
OW 150
|
152 |
+
ET 151
|
153 |
+
▁O 152
|
154 |
+
▁THERE 153
|
155 |
+
▁HO 154
|
156 |
+
ION 155
|
157 |
+
▁WHAT 156
|
158 |
+
▁FE 157
|
159 |
+
▁PA 158
|
160 |
+
US 159
|
161 |
+
MENT 160
|
162 |
+
▁MA 161
|
163 |
+
UT 162
|
164 |
+
▁OUT 163
|
165 |
+
▁THEIR 164
|
166 |
+
▁IF 165
|
167 |
+
▁LI 166
|
168 |
+
▁K 167
|
169 |
+
▁WILL 168
|
170 |
+
▁ARE 169
|
171 |
+
ID 170
|
172 |
+
▁RO 171
|
173 |
+
DE 172
|
174 |
+
TION 173
|
175 |
+
▁WA 174
|
176 |
+
PE 175
|
177 |
+
▁UP 176
|
178 |
+
▁SP 177
|
179 |
+
▁PO 178
|
180 |
+
IGHT 179
|
181 |
+
▁UN 180
|
182 |
+
RU 181
|
183 |
+
▁LO 182
|
184 |
+
AS 183
|
185 |
+
OL 184
|
186 |
+
▁LE 185
|
187 |
+
▁BEEN 186
|
188 |
+
▁SH 187
|
189 |
+
▁RA 188
|
190 |
+
▁SEE 189
|
191 |
+
KE 190
|
192 |
+
UL 191
|
193 |
+
TED 192
|
194 |
+
▁SA 193
|
195 |
+
UN 194
|
196 |
+
UND 195
|
197 |
+
ANT 196
|
198 |
+
▁NE 197
|
199 |
+
IS 198
|
200 |
+
▁THEM 199
|
201 |
+
CI 200
|
202 |
+
GE 201
|
203 |
+
▁COULD 202
|
204 |
+
▁DIS 203
|
205 |
+
OM 204
|
206 |
+
ISH 205
|
207 |
+
HE 206
|
208 |
+
EST 207
|
209 |
+
▁SOME 208
|
210 |
+
ENCE 209
|
211 |
+
ITY 210
|
212 |
+
IVE 211
|
213 |
+
▁US 212
|
214 |
+
▁MORE 213
|
215 |
+
▁EN 214
|
216 |
+
ARD 215
|
217 |
+
ATE 216
|
218 |
+
▁YOUR 217
|
219 |
+
▁INTO 218
|
220 |
+
▁KNOW 219
|
221 |
+
▁CO 220
|
222 |
+
ANCE 221
|
223 |
+
▁TIME 222
|
224 |
+
▁WI 223
|
225 |
+
▁YE 224
|
226 |
+
AGE 225
|
227 |
+
▁NOW 226
|
228 |
+
TI 227
|
229 |
+
FF 228
|
230 |
+
ABLE 229
|
231 |
+
▁VERY 230
|
232 |
+
▁LIKE 231
|
233 |
+
AM 232
|
234 |
+
HI 233
|
235 |
+
Z 234
|
236 |
+
▁OTHER 235
|
237 |
+
▁THAN 236
|
238 |
+
▁LITTLE 237
|
239 |
+
▁DID 238
|
240 |
+
▁LOOK 239
|
241 |
+
TY 240
|
242 |
+
ERS 241
|
243 |
+
▁CAN 242
|
244 |
+
▁CHA 243
|
245 |
+
▁AR 244
|
246 |
+
X 245
|
247 |
+
FUL 246
|
248 |
+
UGH 247
|
249 |
+
▁BA 248
|
250 |
+
▁DAY 249
|
251 |
+
▁ABOUT 250
|
252 |
+
TEN 251
|
253 |
+
IM 252
|
254 |
+
▁ANY 253
|
255 |
+
▁PRE 254
|
256 |
+
▁OVER 255
|
257 |
+
IES 256
|
258 |
+
NESS 257
|
259 |
+
ME 258
|
260 |
+
BLE 259
|
261 |
+
▁M 260
|
262 |
+
ROW 261
|
263 |
+
▁HAS 262
|
264 |
+
▁GREAT 263
|
265 |
+
▁VI 264
|
266 |
+
TA 265
|
267 |
+
▁AFTER 266
|
268 |
+
PER 267
|
269 |
+
▁AGAIN 268
|
270 |
+
HO 269
|
271 |
+
SH 270
|
272 |
+
▁UPON 271
|
273 |
+
▁DI 272
|
274 |
+
▁HAND 273
|
275 |
+
▁COM 274
|
276 |
+
IST 275
|
277 |
+
TURE 276
|
278 |
+
▁STA 277
|
279 |
+
▁THEN 278
|
280 |
+
▁SHOULD 279
|
281 |
+
▁GA 280
|
282 |
+
OUS 281
|
283 |
+
OUR 282
|
284 |
+
▁WELL 283
|
285 |
+
▁ONLY 284
|
286 |
+
MAN 285
|
287 |
+
▁GOOD 286
|
288 |
+
▁TWO 287
|
289 |
+
▁MAR 288
|
290 |
+
▁SAY 289
|
291 |
+
▁HU 290
|
292 |
+
TING 291
|
293 |
+
▁OUR 292
|
294 |
+
RESS 293
|
295 |
+
▁DOWN 294
|
296 |
+
IOUS 295
|
297 |
+
▁BEFORE 296
|
298 |
+
▁DA 297
|
299 |
+
▁NA 298
|
300 |
+
QUI 299
|
301 |
+
▁MADE 300
|
302 |
+
▁EVERY 301
|
303 |
+
▁OLD 302
|
304 |
+
▁EVEN 303
|
305 |
+
IG 304
|
306 |
+
▁COME 305
|
307 |
+
▁GRA 306
|
308 |
+
▁RI 307
|
309 |
+
▁LONG 308
|
310 |
+
OT 309
|
311 |
+
SIDE 310
|
312 |
+
WARD 311
|
313 |
+
▁FO 312
|
314 |
+
▁WHERE 313
|
315 |
+
MO 314
|
316 |
+
LESS 315
|
317 |
+
▁SC 316
|
318 |
+
▁MUST 317
|
319 |
+
▁NEVER 318
|
320 |
+
▁HOW 319
|
321 |
+
▁CAME 320
|
322 |
+
▁SUCH 321
|
323 |
+
▁RU 322
|
324 |
+
▁TAKE 323
|
325 |
+
▁WO 324
|
326 |
+
▁CAR 325
|
327 |
+
UM 326
|
328 |
+
AK 327
|
329 |
+
▁THINK 328
|
330 |
+
▁MUCH 329
|
331 |
+
▁MISTER 330
|
332 |
+
▁MAY 331
|
333 |
+
▁JO 332
|
334 |
+
▁WAY 333
|
335 |
+
▁COMP 334
|
336 |
+
▁THOUGHT 335
|
337 |
+
▁STO 336
|
338 |
+
▁MEN 337
|
339 |
+
▁BACK 338
|
340 |
+
▁DON 339
|
341 |
+
J 340
|
342 |
+
▁LET 341
|
343 |
+
▁TRA 342
|
344 |
+
▁FIRST 343
|
345 |
+
▁JUST 344
|
346 |
+
▁VA 345
|
347 |
+
▁OWN 346
|
348 |
+
▁PLA 347
|
349 |
+
▁MAKE 348
|
350 |
+
ATED 349
|
351 |
+
▁HIMSELF 350
|
352 |
+
▁WENT 351
|
353 |
+
▁PI 352
|
354 |
+
GG 353
|
355 |
+
RING 354
|
356 |
+
▁DU 355
|
357 |
+
▁MIGHT 356
|
358 |
+
▁PART 357
|
359 |
+
▁GIVE 358
|
360 |
+
▁IMP 359
|
361 |
+
▁BU 360
|
362 |
+
▁PER 361
|
363 |
+
▁PLACE 362
|
364 |
+
▁HOUSE 363
|
365 |
+
▁THROUGH 364
|
366 |
+
IAN 365
|
367 |
+
▁SW 366
|
368 |
+
▁UNDER 367
|
369 |
+
QUE 368
|
370 |
+
▁AWAY 369
|
371 |
+
▁LOVE 370
|
372 |
+
QUA 371
|
373 |
+
▁LIFE 372
|
374 |
+
▁GET 373
|
375 |
+
▁WITHOUT 374
|
376 |
+
▁PASS 375
|
377 |
+
▁TURN 376
|
378 |
+
IGN 377
|
379 |
+
▁HEAD 378
|
380 |
+
▁MOST 379
|
381 |
+
▁THOSE 380
|
382 |
+
▁SHALL 381
|
383 |
+
▁EYES 382
|
384 |
+
▁COL 383
|
385 |
+
▁STILL 384
|
386 |
+
▁NIGHT 385
|
387 |
+
▁NOTHING 386
|
388 |
+
ITION 387
|
389 |
+
HA 388
|
390 |
+
▁TELL 389
|
391 |
+
▁WORK 390
|
392 |
+
▁LAST 391
|
393 |
+
▁NEW 392
|
394 |
+
▁FACE 393
|
395 |
+
▁HI 394
|
396 |
+
▁WORD 395
|
397 |
+
▁FOUND 396
|
398 |
+
▁COUNT 397
|
399 |
+
▁OB 398
|
400 |
+
▁WHILE 399
|
401 |
+
▁SHA 400
|
402 |
+
▁MEAN 401
|
403 |
+
▁SAW 402
|
404 |
+
▁PEOPLE 403
|
405 |
+
▁FRIEND 404
|
406 |
+
▁THREE 405
|
407 |
+
▁ROOM 406
|
408 |
+
▁SAME 407
|
409 |
+
▁THOUGH 408
|
410 |
+
▁RIGHT 409
|
411 |
+
▁CHILD 410
|
412 |
+
▁FATHER 411
|
413 |
+
▁ANOTHER 412
|
414 |
+
▁HEART 413
|
415 |
+
▁WANT 414
|
416 |
+
▁TOOK 415
|
417 |
+
OOK 416
|
418 |
+
▁LIGHT 417
|
419 |
+
▁MISSUS 418
|
420 |
+
▁OPEN 419
|
421 |
+
▁JU 420
|
422 |
+
▁ASKED 421
|
423 |
+
PORT 422
|
424 |
+
▁LEFT 423
|
425 |
+
▁JA 424
|
426 |
+
▁WORLD 425
|
427 |
+
▁HOME 426
|
428 |
+
▁WHY 427
|
429 |
+
▁ALWAYS 428
|
430 |
+
▁ANSWER 429
|
431 |
+
▁SEEMED 430
|
432 |
+
▁SOMETHING 431
|
433 |
+
▁GIRL 432
|
434 |
+
▁BECAUSE 433
|
435 |
+
▁NAME 434
|
436 |
+
▁TOLD 435
|
437 |
+
▁NI 436
|
438 |
+
▁HIGH 437
|
439 |
+
IZE 438
|
440 |
+
▁WOMAN 439
|
441 |
+
▁FOLLOW 440
|
442 |
+
▁RETURN 441
|
443 |
+
▁KNEW 442
|
444 |
+
▁EACH 443
|
445 |
+
▁KIND 444
|
446 |
+
▁JE 445
|
447 |
+
▁ACT 446
|
448 |
+
▁LU 447
|
449 |
+
▁CERTAIN 448
|
450 |
+
▁YEARS 449
|
451 |
+
▁QUITE 450
|
452 |
+
▁APPEAR 451
|
453 |
+
▁BETTER 452
|
454 |
+
▁HALF 453
|
455 |
+
▁PRESENT 454
|
456 |
+
▁PRINCE 455
|
457 |
+
SHIP 456
|
458 |
+
▁ALSO 457
|
459 |
+
▁BEGAN 458
|
460 |
+
▁HAVING 459
|
461 |
+
▁ENOUGH 460
|
462 |
+
▁PERSON 461
|
463 |
+
▁LADY 462
|
464 |
+
▁WHITE 463
|
465 |
+
▁COURSE 464
|
466 |
+
▁VOICE 465
|
467 |
+
▁SPEAK 466
|
468 |
+
▁POWER 467
|
469 |
+
▁MORNING 468
|
470 |
+
▁BETWEEN 469
|
471 |
+
▁AMONG 470
|
472 |
+
▁KEEP 471
|
473 |
+
▁WALK 472
|
474 |
+
▁MATTER 473
|
475 |
+
▁TEA 474
|
476 |
+
▁BELIEVE 475
|
477 |
+
▁SMALL 476
|
478 |
+
▁TALK 477
|
479 |
+
▁FELT 478
|
480 |
+
▁HORSE 479
|
481 |
+
▁MYSELF 480
|
482 |
+
▁SIX 481
|
483 |
+
▁HOWEVER 482
|
484 |
+
▁FULL 483
|
485 |
+
▁HERSELF 484
|
486 |
+
▁POINT 485
|
487 |
+
▁STOOD 486
|
488 |
+
▁HUNDRED 487
|
489 |
+
▁ALMOST 488
|
490 |
+
▁SINCE 489
|
491 |
+
▁LARGE 490
|
492 |
+
▁LEAVE 491
|
493 |
+
▁PERHAPS 492
|
494 |
+
▁DARK 493
|
495 |
+
▁SUDDEN 494
|
496 |
+
▁REPLIED 495
|
497 |
+
▁ANYTHING 496
|
498 |
+
▁WONDER 497
|
499 |
+
▁UNTIL 498
|
500 |
+
Q 499
|
501 |
+
#0 500
|
502 |
+
#1 501
|
data/lang_bpe_500/words.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
data/lm/G_4_gram.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c91581bc632f9c72c557ccdf726298255b9627a6ac38270c51891459b82630e9
|
3 |
+
size 3700956587
|
decoding_results/1best/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/1best/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/1best/log-decode-epoch-40-avg-16-use-averaged-model-2023-06-13-19-45-46
ADDED
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
2023-06-13 19:45:46,339 INFO [ctc_decode.py:641] Decoding started
|
2 |
+
2023-06-13 19:45:46,340 INFO [ctc_decode.py:647] Device: cuda:0
|
3 |
+
2023-06-13 19:45:46,340 INFO [ctc_decode.py:648] {'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 50, 'reset_interval': 200, 'valid_interval': 3000, 'feature_dim': 80, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'c51a0b9684442a88ee37f3ce0af686a04b66855b', 'k2-git-date': 'Mon May 1 21:38:03 2023', 'lhotse-version': '1.12.0.dev+git.891bad1.clean', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'new-zipformer-add-ctc', 'icefall-git-sha1': '046b6cb6-dirty', 'icefall-git-date': 'Fri Jun 2 15:51:49 2023', 'icefall-path': '/ceph-zw/workspace/zipformer/icefall_zipformer', 'k2-path': '/ceph-zw/workspace/k2/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-zw/workspace/share/lhotse/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-1-1220091118-57c4d55446-mlpzc', 'IP address': '10.177.22.19'}, 'frame_shift_ms': 10, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'use_double_scores': True, 'epoch': 40, 'iter': 0, 'avg': 16, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp-ctc-rnnt'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'context_size': 2, 'decoding_method': '1best', 'num_paths': 100, 'nbest_scale': 1.0, 'hlg_scale': 0.6, 'lm_dir': PosixPath('data/my_lm'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'decoder_dim': 512, 'joiner_dim': 512, 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'use_transducer': True, 'use_ctc': True, 'full_libri': True, 'mini_libri': False, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 300, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'drop_last': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp-ctc-rnnt/1best'), 'suffix': 'epoch-40-avg-16-use-averaged-model'}
|
4 |
+
2023-06-13 19:45:46,609 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
|
5 |
+
2023-06-13 19:45:51,497 INFO [ctc_decode.py:726] About to create model
|
6 |
+
2023-06-13 19:45:52,095 INFO [ctc_decode.py:793] Calculating the averaged model over epoch range from 24 (excluded) to 40
|
7 |
+
2023-06-13 19:45:54,495 INFO [ctc_decode.py:810] Number of model parameters: 65805511
|
8 |
+
2023-06-13 19:45:54,496 INFO [asr_datamodule.py:465] About to get test-clean cuts
|
9 |
+
2023-06-13 19:45:54,500 INFO [asr_datamodule.py:472] About to get test-other cuts
|
10 |
+
2023-06-13 19:45:55,448 INFO [ctc_decode.py:558] batch 0/?, cuts processed until now is 21
|
11 |
+
2023-06-13 19:46:22,063 INFO [ctc_decode.py:572] The transcripts are stored in zipformer/exp-ctc-rnnt/1best/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
12 |
+
2023-06-13 19:46:22,133 INFO [utils.py:561] [test-clean-no_rescore] %WER 2.46% [1294 / 52576, 181 ins, 93 del, 1020 sub ]
|
13 |
+
2023-06-13 19:46:22,308 INFO [ctc_decode.py:581] Wrote detailed error stats to zipformer/exp-ctc-rnnt/1best/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
14 |
+
2023-06-13 19:46:22,309 INFO [ctc_decode.py:595]
|
15 |
+
For test-clean, WER of different settings are:
|
16 |
+
no_rescore 2.46 best for test-clean
|
17 |
+
|
18 |
+
2023-06-13 19:46:23,023 INFO [ctc_decode.py:558] batch 0/?, cuts processed until now is 26
|
19 |
+
2023-06-13 19:46:50,469 INFO [ctc_decode.py:572] The transcripts are stored in zipformer/exp-ctc-rnnt/1best/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
20 |
+
2023-06-13 19:46:50,541 INFO [utils.py:561] [test-other-no_rescore] %WER 5.11% [2674 / 52343, 288 ins, 256 del, 2130 sub ]
|
21 |
+
2023-06-13 19:46:50,716 INFO [ctc_decode.py:581] Wrote detailed error stats to zipformer/exp-ctc-rnnt/1best/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
22 |
+
2023-06-13 19:46:50,717 INFO [ctc_decode.py:595]
|
23 |
+
For test-other, WER of different settings are:
|
24 |
+
no_rescore 5.11 best for test-other
|
25 |
+
|
26 |
+
2023-06-13 19:46:50,717 INFO [ctc_decode.py:843] Done!
|
decoding_results/1best/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/1best/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/1best/wer-summary-test-clean-epoch-40-avg-16-use-averaged-model.txt
ADDED
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
1 |
+
settings WER
|
2 |
+
no_rescore 2.46
|
decoding_results/1best/wer-summary-test-other-epoch-40-avg-16-use-averaged-model.txt
ADDED
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
1 |
+
settings WER
|
2 |
+
no_rescore 5.11
|
decoding_results/ctc-decoding/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/ctc-decoding/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/ctc-decoding/log-decode-epoch-40-avg-16-use-averaged-model-2023-06-13-19-42-20
ADDED
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
2023-06-13 19:42:20,348 INFO [ctc_decode.py:641] Decoding started
|
2 |
+
2023-06-13 19:42:20,349 INFO [ctc_decode.py:647] Device: cuda:0
|
3 |
+
2023-06-13 19:42:20,349 INFO [ctc_decode.py:648] {'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 50, 'reset_interval': 200, 'valid_interval': 3000, 'feature_dim': 80, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'c51a0b9684442a88ee37f3ce0af686a04b66855b', 'k2-git-date': 'Mon May 1 21:38:03 2023', 'lhotse-version': '1.12.0.dev+git.891bad1.clean', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'new-zipformer-add-ctc', 'icefall-git-sha1': '046b6cb6-dirty', 'icefall-git-date': 'Fri Jun 2 15:51:49 2023', 'icefall-path': '/ceph-zw/workspace/zipformer/icefall_zipformer', 'k2-path': '/ceph-zw/workspace/k2/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-zw/workspace/share/lhotse/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-1-1220091118-57c4d55446-mlpzc', 'IP address': '10.177.22.19'}, 'frame_shift_ms': 10, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'use_double_scores': True, 'epoch': 40, 'iter': 0, 'avg': 16, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp-ctc-rnnt'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'context_size': 2, 'decoding_method': 'ctc-decoding', 'num_paths': 100, 'nbest_scale': 1.0, 'hlg_scale': 0.6, 'lm_dir': PosixPath('data/my_lm'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'decoder_dim': 512, 'joiner_dim': 512, 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'use_transducer': True, 'use_ctc': True, 'full_libri': True, 'mini_libri': False, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 300, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'drop_last': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp-ctc-rnnt/ctc-decoding'), 'suffix': 'epoch-40-avg-16-use-averaged-model'}
|
4 |
+
2023-06-13 19:42:20,615 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
|
5 |
+
2023-06-13 19:42:24,473 INFO [ctc_decode.py:726] About to create model
|
6 |
+
2023-06-13 19:42:25,063 INFO [ctc_decode.py:793] Calculating the averaged model over epoch range from 24 (excluded) to 40
|
7 |
+
2023-06-13 19:42:27,461 INFO [ctc_decode.py:810] Number of model parameters: 65805511
|
8 |
+
2023-06-13 19:42:27,462 INFO [asr_datamodule.py:465] About to get test-clean cuts
|
9 |
+
2023-06-13 19:42:27,466 INFO [asr_datamodule.py:472] About to get test-other cuts
|
10 |
+
2023-06-13 19:42:28,473 INFO [ctc_decode.py:558] batch 0/?, cuts processed until now is 21
|
11 |
+
2023-06-13 19:42:54,474 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([4.6098, 4.0538, 4.2424, 4.1121], device='cuda:0')
|
12 |
+
2023-06-13 19:42:57,548 INFO [ctc_decode.py:572] The transcripts are stored in zipformer/exp-ctc-rnnt/ctc-decoding/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
13 |
+
2023-06-13 19:42:57,621 INFO [utils.py:561] [test-clean-ctc-decoding] %WER 2.40% [1262 / 52576, 133 ins, 89 del, 1040 sub ]
|
14 |
+
2023-06-13 19:42:57,796 INFO [ctc_decode.py:581] Wrote detailed error stats to zipformer/exp-ctc-rnnt/ctc-decoding/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
15 |
+
2023-06-13 19:42:57,796 INFO [ctc_decode.py:595]
|
16 |
+
For test-clean, WER of different settings are:
|
17 |
+
ctc-decoding 2.4 best for test-clean
|
18 |
+
|
19 |
+
2023-06-13 19:42:58,556 INFO [ctc_decode.py:558] batch 0/?, cuts processed until now is 26
|
20 |
+
2023-06-13 19:43:06,459 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([4.0038, 3.6233, 3.5348, 3.5192], device='cuda:0')
|
21 |
+
2023-06-13 19:43:10,237 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([4.2716, 2.5148, 3.4661, 1.9089], device='cuda:0')
|
22 |
+
2023-06-13 19:43:19,177 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([2.3797, 3.3571, 3.0888, 2.5630], device='cuda:0')
|
23 |
+
2023-06-13 19:43:25,600 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([4.5188, 4.0134, 3.8820, 4.2079], device='cuda:0')
|
24 |
+
2023-06-13 19:43:27,974 INFO [ctc_decode.py:572] The transcripts are stored in zipformer/exp-ctc-rnnt/ctc-decoding/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
25 |
+
2023-06-13 19:43:28,074 INFO [utils.py:561] [test-other-ctc-decoding] %WER 5.66% [2961 / 52343, 278 ins, 224 del, 2459 sub ]
|
26 |
+
2023-06-13 19:43:28,356 INFO [ctc_decode.py:581] Wrote detailed error stats to zipformer/exp-ctc-rnnt/ctc-decoding/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
27 |
+
2023-06-13 19:43:28,357 INFO [ctc_decode.py:595]
|
28 |
+
For test-other, WER of different settings are:
|
29 |
+
ctc-decoding 5.66 best for test-other
|
30 |
+
|
31 |
+
2023-06-13 19:43:28,357 INFO [ctc_decode.py:843] Done!
|
decoding_results/ctc-decoding/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/ctc-decoding/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/ctc-decoding/wer-summary-test-clean-epoch-40-avg-16-use-averaged-model.txt
ADDED
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
1 |
+
settings WER
|
2 |
+
ctc-decoding 2.4
|
decoding_results/ctc-decoding/wer-summary-test-other-epoch-40-avg-16-use-averaged-model.txt
ADDED
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
1 |
+
settings WER
|
2 |
+
ctc-decoding 5.66
|
decoding_results/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/nbest-rescoring/log-decode-epoch-40-avg-16-use-averaged-model-2023-06-13-17-24-42
ADDED
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
2023-06-13 17:24:42,946 INFO [ctc_decode.py:633] Decoding started
|
2 |
+
2023-06-13 17:24:42,946 INFO [ctc_decode.py:639] Device: cuda:0
|
3 |
+
2023-06-13 17:24:42,946 INFO [ctc_decode.py:640] {'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 50, 'reset_interval': 200, 'valid_interval': 3000, 'feature_dim': 80, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'c51a0b9684442a88ee37f3ce0af686a04b66855b', 'k2-git-date': 'Mon May 1 21:38:03 2023', 'lhotse-version': '1.12.0.dev+git.891bad1.clean', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'new-zipformer-add-ctc', 'icefall-git-sha1': '046b6cb6-clean', 'icefall-git-date': 'Fri Jun 2 15:51:49 2023', 'icefall-path': '/ceph-zw/workspace/zipformer/icefall_zipformer', 'k2-path': '/ceph-zw/workspace/k2/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-zw/workspace/share/lhotse/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-1-1220091118-57c4d55446-mlpzc', 'IP address': '10.177.22.19'}, 'frame_shift_ms': 10, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'use_double_scores': True, 'epoch': 40, 'iter': 0, 'avg': 16, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp-ctc-rnnt'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'context_size': 2, 'decoding_method': 'nbest-rescoring', 'num_paths': 100, 'nbest_scale': 1.0, 'hlg_scale': 0.6, 'lm_dir': PosixPath('data/lm'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'decoder_dim': 512, 'joiner_dim': 512, 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'use_transducer': True, 'use_ctc': True, 'full_libri': True, 'mini_libri': False, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 300, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'drop_last': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp-ctc-rnnt/nbest-rescoring'), 'suffix': 'epoch-40-avg-16-use-averaged-model'}
|
4 |
+
2023-06-13 17:24:43,219 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
|
5 |
+
2023-06-13 17:24:48,038 INFO [ctc_decode.py:676] Loading G_4_gram.fst.txt
|
6 |
+
2023-06-13 17:24:48,038 WARNING [ctc_decode.py:677] It may take 8 minutes.
|
decoding_results/nbest-rescoring/log-decode-epoch-40-avg-16-use-averaged-model-2023-06-13-17-28-30
ADDED
@@ -0,0 +1,180 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
2023-06-13 17:28:30,904 INFO [ctc_decode.py:633] Decoding started
|
2 |
+
2023-06-13 17:28:30,904 INFO [ctc_decode.py:639] Device: cuda:0
|
3 |
+
2023-06-13 17:28:30,904 INFO [ctc_decode.py:640] {'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 50, 'reset_interval': 200, 'valid_interval': 3000, 'feature_dim': 80, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'c51a0b9684442a88ee37f3ce0af686a04b66855b', 'k2-git-date': 'Mon May 1 21:38:03 2023', 'lhotse-version': '1.12.0.dev+git.891bad1.clean', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'new-zipformer-add-ctc', 'icefall-git-sha1': '046b6cb6-clean', 'icefall-git-date': 'Fri Jun 2 15:51:49 2023', 'icefall-path': '/ceph-zw/workspace/zipformer/icefall_zipformer', 'k2-path': '/ceph-zw/workspace/k2/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-zw/workspace/share/lhotse/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-1-1220091118-57c4d55446-mlpzc', 'IP address': '10.177.22.19'}, 'frame_shift_ms': 10, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'use_double_scores': True, 'epoch': 40, 'iter': 0, 'avg': 16, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp-ctc-rnnt'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'context_size': 2, 'decoding_method': 'nbest-rescoring', 'num_paths': 100, 'nbest_scale': 1.0, 'hlg_scale': 0.6, 'lm_dir': PosixPath('data/my_lm'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'decoder_dim': 512, 'joiner_dim': 512, 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'use_transducer': True, 'use_ctc': True, 'full_libri': True, 'mini_libri': False, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 300, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'drop_last': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp-ctc-rnnt/nbest-rescoring'), 'suffix': 'epoch-40-avg-16-use-averaged-model'}
|
4 |
+
2023-06-13 17:28:31,179 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
|
5 |
+
2023-06-13 17:28:35,887 INFO [ctc_decode.py:701] Loading pre-compiled G_4_gram.pt
|
6 |
+
2023-06-13 17:28:58,167 INFO [ctc_decode.py:718] About to create model
|
7 |
+
2023-06-13 17:28:58,818 INFO [ctc_decode.py:785] Calculating the averaged model over epoch range from 24 (excluded) to 40
|
8 |
+
2023-06-13 17:29:02,258 INFO [ctc_decode.py:802] Number of model parameters: 65805511
|
9 |
+
2023-06-13 17:29:02,259 INFO [asr_datamodule.py:465] About to get test-clean cuts
|
10 |
+
2023-06-13 17:29:02,262 INFO [asr_datamodule.py:472] About to get test-other cuts
|
11 |
+
2023-06-13 17:29:03,502 INFO [ctc_decode.py:550] batch 0/?, cuts processed until now is 21
|
12 |
+
2023-06-13 17:29:39,698 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([4.6109, 4.3406, 4.1379, 4.6409], device='cuda:0')
|
13 |
+
2023-06-13 17:29:53,312 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
14 |
+
2023-06-13 17:29:53,387 INFO [utils.py:561] [test-clean-lm_scale_0.1] %WER 2.49% [1309 / 52576, 219 ins, 68 del, 1022 sub ]
|
15 |
+
2023-06-13 17:29:53,562 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
16 |
+
2023-06-13 17:29:53,585 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
17 |
+
2023-06-13 17:29:53,654 INFO [utils.py:561] [test-clean-lm_scale_0.2] %WER 2.45% [1289 / 52576, 211 ins, 68 del, 1010 sub ]
|
18 |
+
2023-06-13 17:29:53,833 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
19 |
+
2023-06-13 17:29:53,858 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
20 |
+
2023-06-13 17:29:54,096 INFO [utils.py:561] [test-clean-lm_scale_0.3] %WER 2.42% [1272 / 52576, 202 ins, 72 del, 998 sub ]
|
21 |
+
2023-06-13 17:29:54,266 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
22 |
+
2023-06-13 17:29:54,289 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
23 |
+
2023-06-13 17:29:54,356 INFO [utils.py:561] [test-clean-lm_scale_0.4] %WER 2.39% [1254 / 52576, 192 ins, 74 del, 988 sub ]
|
24 |
+
2023-06-13 17:29:54,529 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
25 |
+
2023-06-13 17:29:54,552 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
26 |
+
2023-06-13 17:29:54,622 INFO [utils.py:561] [test-clean-lm_scale_0.5] %WER 2.37% [1248 / 52576, 184 ins, 81 del, 983 sub ]
|
27 |
+
2023-06-13 17:29:54,795 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
28 |
+
2023-06-13 17:29:54,817 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
29 |
+
2023-06-13 17:29:54,886 INFO [utils.py:561] [test-clean-lm_scale_0.6] %WER 2.38% [1250 / 52576, 178 ins, 87 del, 985 sub ]
|
30 |
+
2023-06-13 17:29:55,057 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
31 |
+
2023-06-13 17:29:55,080 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
32 |
+
2023-06-13 17:29:55,147 INFO [utils.py:561] [test-clean-lm_scale_0.7] %WER 2.40% [1261 / 52576, 172 ins, 97 del, 992 sub ]
|
33 |
+
2023-06-13 17:29:55,320 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
34 |
+
2023-06-13 17:29:55,342 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
35 |
+
2023-06-13 17:29:55,409 INFO [utils.py:561] [test-clean-lm_scale_0.8] %WER 2.42% [1272 / 52576, 167 ins, 108 del, 997 sub ]
|
36 |
+
2023-06-13 17:29:55,577 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
37 |
+
2023-06-13 17:29:55,599 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
38 |
+
2023-06-13 17:29:55,822 INFO [utils.py:561] [test-clean-lm_scale_0.9] %WER 2.44% [1284 / 52576, 158 ins, 123 del, 1003 sub ]
|
39 |
+
2023-06-13 17:29:55,994 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
40 |
+
2023-06-13 17:29:56,016 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
41 |
+
2023-06-13 17:29:56,082 INFO [utils.py:561] [test-clean-lm_scale_1.0] %WER 2.50% [1312 / 52576, 154 ins, 140 del, 1018 sub ]
|
42 |
+
2023-06-13 17:29:56,252 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
43 |
+
2023-06-13 17:29:56,275 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
44 |
+
2023-06-13 17:29:56,343 INFO [utils.py:561] [test-clean-lm_scale_1.1] %WER 2.56% [1348 / 52576, 156 ins, 155 del, 1037 sub ]
|
45 |
+
2023-06-13 17:29:56,511 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
46 |
+
2023-06-13 17:29:56,534 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
47 |
+
2023-06-13 17:29:56,601 INFO [utils.py:561] [test-clean-lm_scale_1.2] %WER 2.60% [1367 / 52576, 153 ins, 162 del, 1052 sub ]
|
48 |
+
2023-06-13 17:29:56,773 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
49 |
+
2023-06-13 17:29:56,795 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
50 |
+
2023-06-13 17:29:56,864 INFO [utils.py:561] [test-clean-lm_scale_1.3] %WER 2.66% [1396 / 52576, 153 ins, 179 del, 1064 sub ]
|
51 |
+
2023-06-13 17:29:57,033 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
52 |
+
2023-06-13 17:29:57,056 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
53 |
+
2023-06-13 17:29:57,123 INFO [utils.py:561] [test-clean-lm_scale_1.4] %WER 2.69% [1414 / 52576, 150 ins, 191 del, 1073 sub ]
|
54 |
+
2023-06-13 17:29:57,294 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
55 |
+
2023-06-13 17:29:57,316 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
56 |
+
2023-06-13 17:29:57,383 INFO [utils.py:561] [test-clean-lm_scale_1.5] %WER 2.73% [1436 / 52576, 149 ins, 203 del, 1084 sub ]
|
57 |
+
2023-06-13 17:29:57,707 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
58 |
+
2023-06-13 17:29:57,730 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
59 |
+
2023-06-13 17:29:57,798 INFO [utils.py:561] [test-clean-lm_scale_1.6] %WER 2.80% [1471 / 52576, 149 ins, 220 del, 1102 sub ]
|
60 |
+
2023-06-13 17:29:57,974 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
61 |
+
2023-06-13 17:29:57,996 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
62 |
+
2023-06-13 17:29:58,062 INFO [utils.py:561] [test-clean-lm_scale_1.7] %WER 2.85% [1501 / 52576, 149 ins, 227 del, 1125 sub ]
|
63 |
+
2023-06-13 17:29:58,237 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
64 |
+
2023-06-13 17:29:58,259 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
65 |
+
2023-06-13 17:29:58,326 INFO [utils.py:561] [test-clean-lm_scale_1.8] %WER 2.89% [1520 / 52576, 149 ins, 233 del, 1138 sub ]
|
66 |
+
2023-06-13 17:29:58,494 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
67 |
+
2023-06-13 17:29:58,516 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
68 |
+
2023-06-13 17:29:58,583 INFO [utils.py:561] [test-clean-lm_scale_1.9] %WER 2.94% [1545 / 52576, 147 ins, 242 del, 1156 sub ]
|
69 |
+
2023-06-13 17:29:58,750 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
70 |
+
2023-06-13 17:29:58,773 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
71 |
+
2023-06-13 17:29:58,841 INFO [utils.py:561] [test-clean-lm_scale_2.0] %WER 2.98% [1566 / 52576, 147 ins, 252 del, 1167 sub ]
|
72 |
+
2023-06-13 17:29:59,014 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
73 |
+
2023-06-13 17:29:59,015 INFO [ctc_decode.py:587]
|
74 |
+
For test-clean, WER of different settings are:
|
75 |
+
lm_scale_0.5 2.37 best for test-clean
|
76 |
+
lm_scale_0.6 2.38
|
77 |
+
lm_scale_0.4 2.39
|
78 |
+
lm_scale_0.7 2.4
|
79 |
+
lm_scale_0.3 2.42
|
80 |
+
lm_scale_0.8 2.42
|
81 |
+
lm_scale_0.9 2.44
|
82 |
+
lm_scale_0.2 2.45
|
83 |
+
lm_scale_0.1 2.49
|
84 |
+
lm_scale_1.0 2.5
|
85 |
+
lm_scale_1.1 2.56
|
86 |
+
lm_scale_1.2 2.6
|
87 |
+
lm_scale_1.3 2.66
|
88 |
+
lm_scale_1.4 2.69
|
89 |
+
lm_scale_1.5 2.73
|
90 |
+
lm_scale_1.6 2.8
|
91 |
+
lm_scale_1.7 2.85
|
92 |
+
lm_scale_1.8 2.89
|
93 |
+
lm_scale_1.9 2.94
|
94 |
+
lm_scale_2.0 2.98
|
95 |
+
|
96 |
+
2023-06-13 17:30:00,580 INFO [ctc_decode.py:550] batch 0/?, cuts processed until now is 26
|
97 |
+
2023-06-13 17:31:00,699 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
98 |
+
2023-06-13 17:31:00,770 INFO [utils.py:561] [test-other-lm_scale_0.1] %WER 5.20% [2721 / 52343, 362 ins, 183 del, 2176 sub ]
|
99 |
+
2023-06-13 17:31:00,946 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
100 |
+
2023-06-13 17:31:00,970 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
101 |
+
2023-06-13 17:31:01,039 INFO [utils.py:561] [test-other-lm_scale_0.2] %WER 5.10% [2670 / 52343, 355 ins, 183 del, 2132 sub ]
|
102 |
+
2023-06-13 17:31:01,212 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
103 |
+
2023-06-13 17:31:01,235 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
104 |
+
2023-06-13 17:31:01,306 INFO [utils.py:561] [test-other-lm_scale_0.3] %WER 5.04% [2638 / 52343, 342 ins, 187 del, 2109 sub ]
|
105 |
+
2023-06-13 17:31:01,481 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
106 |
+
2023-06-13 17:31:01,503 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
107 |
+
2023-06-13 17:31:01,573 INFO [utils.py:561] [test-other-lm_scale_0.4] %WER 4.99% [2610 / 52343, 323 ins, 198 del, 2089 sub ]
|
108 |
+
2023-06-13 17:31:01,745 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
109 |
+
2023-06-13 17:31:01,769 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
110 |
+
2023-06-13 17:31:01,842 INFO [utils.py:561] [test-other-lm_scale_0.5] %WER 4.95% [2592 / 52343, 316 ins, 214 del, 2062 sub ]
|
111 |
+
2023-06-13 17:31:02,020 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
112 |
+
2023-06-13 17:31:02,042 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
113 |
+
2023-06-13 17:31:02,110 INFO [utils.py:561] [test-other-lm_scale_0.6] %WER 4.93% [2583 / 52343, 290 ins, 234 del, 2059 sub ]
|
114 |
+
2023-06-13 17:31:02,283 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
115 |
+
2023-06-13 17:31:02,306 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
116 |
+
2023-06-13 17:31:02,378 INFO [utils.py:561] [test-other-lm_scale_0.7] %WER 4.94% [2584 / 52343, 279 ins, 250 del, 2055 sub ]
|
117 |
+
2023-06-13 17:31:02,550 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
118 |
+
2023-06-13 17:31:02,573 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
119 |
+
2023-06-13 17:31:02,641 INFO [utils.py:561] [test-other-lm_scale_0.8] %WER 4.99% [2614 / 52343, 265 ins, 288 del, 2061 sub ]
|
120 |
+
2023-06-13 17:31:02,975 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
121 |
+
2023-06-13 17:31:02,998 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
122 |
+
2023-06-13 17:31:03,067 INFO [utils.py:561] [test-other-lm_scale_0.9] %WER 5.03% [2635 / 52343, 252 ins, 305 del, 2078 sub ]
|
123 |
+
2023-06-13 17:31:03,244 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
124 |
+
2023-06-13 17:31:03,268 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
125 |
+
2023-06-13 17:31:03,338 INFO [utils.py:561] [test-other-lm_scale_1.0] %WER 5.08% [2660 / 52343, 250 ins, 331 del, 2079 sub ]
|
126 |
+
2023-06-13 17:31:03,513 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
127 |
+
2023-06-13 17:31:03,537 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
128 |
+
2023-06-13 17:31:03,608 INFO [utils.py:561] [test-other-lm_scale_1.1] %WER 5.16% [2701 / 52343, 244 ins, 350 del, 2107 sub ]
|
129 |
+
2023-06-13 17:31:03,782 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
130 |
+
2023-06-13 17:31:03,805 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
131 |
+
2023-06-13 17:31:03,874 INFO [utils.py:561] [test-other-lm_scale_1.2] %WER 5.25% [2748 / 52343, 241 ins, 383 del, 2124 sub ]
|
132 |
+
2023-06-13 17:31:04,060 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
133 |
+
2023-06-13 17:31:04,085 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
134 |
+
2023-06-13 17:31:04,153 INFO [utils.py:561] [test-other-lm_scale_1.3] %WER 5.30% [2772 / 52343, 237 ins, 404 del, 2131 sub ]
|
135 |
+
2023-06-13 17:31:04,327 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
136 |
+
2023-06-13 17:31:04,350 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
137 |
+
2023-06-13 17:31:04,418 INFO [utils.py:561] [test-other-lm_scale_1.4] %WER 5.38% [2815 / 52343, 231 ins, 422 del, 2162 sub ]
|
138 |
+
2023-06-13 17:31:04,591 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
139 |
+
2023-06-13 17:31:04,615 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
140 |
+
2023-06-13 17:31:04,683 INFO [utils.py:561] [test-other-lm_scale_1.5] %WER 5.45% [2855 / 52343, 230 ins, 452 del, 2173 sub ]
|
141 |
+
2023-06-13 17:31:05,037 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
142 |
+
2023-06-13 17:31:05,060 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
143 |
+
2023-06-13 17:31:05,129 INFO [utils.py:561] [test-other-lm_scale_1.6] %WER 5.52% [2887 / 52343, 226 ins, 469 del, 2192 sub ]
|
144 |
+
2023-06-13 17:31:05,304 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
145 |
+
2023-06-13 17:31:05,327 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
146 |
+
2023-06-13 17:31:05,395 INFO [utils.py:561] [test-other-lm_scale_1.7] %WER 5.58% [2919 / 52343, 227 ins, 492 del, 2200 sub ]
|
147 |
+
2023-06-13 17:31:05,572 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
148 |
+
2023-06-13 17:31:05,595 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
149 |
+
2023-06-13 17:31:05,664 INFO [utils.py:561] [test-other-lm_scale_1.8] %WER 5.63% [2947 / 52343, 227 ins, 506 del, 2214 sub ]
|
150 |
+
2023-06-13 17:31:05,840 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
151 |
+
2023-06-13 17:31:05,863 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
152 |
+
2023-06-13 17:31:05,938 INFO [utils.py:561] [test-other-lm_scale_1.9] %WER 5.70% [2984 / 52343, 227 ins, 520 del, 2237 sub ]
|
153 |
+
2023-06-13 17:31:06,113 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
154 |
+
2023-06-13 17:31:06,136 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
155 |
+
2023-06-13 17:31:06,204 INFO [utils.py:561] [test-other-lm_scale_2.0] %WER 5.74% [3007 / 52343, 224 ins, 532 del, 2251 sub ]
|
156 |
+
2023-06-13 17:31:06,379 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
157 |
+
2023-06-13 17:31:06,380 INFO [ctc_decode.py:587]
|
158 |
+
For test-other, WER of different settings are:
|
159 |
+
lm_scale_0.6 4.93 best for test-other
|
160 |
+
lm_scale_0.7 4.94
|
161 |
+
lm_scale_0.5 4.95
|
162 |
+
lm_scale_0.4 4.99
|
163 |
+
lm_scale_0.8 4.99
|
164 |
+
lm_scale_0.9 5.03
|
165 |
+
lm_scale_0.3 5.04
|
166 |
+
lm_scale_1.0 5.08
|
167 |
+
lm_scale_0.2 5.1
|
168 |
+
lm_scale_1.1 5.16
|
169 |
+
lm_scale_0.1 5.2
|
170 |
+
lm_scale_1.2 5.25
|
171 |
+
lm_scale_1.3 5.3
|
172 |
+
lm_scale_1.4 5.38
|
173 |
+
lm_scale_1.5 5.45
|
174 |
+
lm_scale_1.6 5.52
|
175 |
+
lm_scale_1.7 5.58
|
176 |
+
lm_scale_1.8 5.63
|
177 |
+
lm_scale_1.9 5.7
|
178 |
+
lm_scale_2.0 5.74
|
179 |
+
|
180 |
+
2023-06-13 17:31:06,380 INFO [ctc_decode.py:835] Done!
|
decoding_results/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/nbest-rescoring/wer-summary-test-clean-epoch-40-avg-16-use-averaged-model.txt
ADDED
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
settings WER
|
2 |
+
lm_scale_0.5 2.37
|
3 |
+
lm_scale_0.6 2.38
|
4 |
+
lm_scale_0.4 2.39
|
5 |
+
lm_scale_0.7 2.4
|
6 |
+
lm_scale_0.3 2.42
|
7 |
+
lm_scale_0.8 2.42
|
8 |
+
lm_scale_0.9 2.44
|
9 |
+
lm_scale_0.2 2.45
|
10 |
+
lm_scale_0.1 2.49
|
11 |
+
lm_scale_1.0 2.5
|
12 |
+
lm_scale_1.1 2.56
|
13 |
+
lm_scale_1.2 2.6
|
14 |
+
lm_scale_1.3 2.66
|
15 |
+
lm_scale_1.4 2.69
|
16 |
+
lm_scale_1.5 2.73
|
17 |
+
lm_scale_1.6 2.8
|
18 |
+
lm_scale_1.7 2.85
|
19 |
+
lm_scale_1.8 2.89
|
20 |
+
lm_scale_1.9 2.94
|
21 |
+
lm_scale_2.0 2.98
|
decoding_results/nbest-rescoring/wer-summary-test-other-epoch-40-avg-16-use-averaged-model.txt
ADDED
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
settings WER
|
2 |
+
lm_scale_0.6 4.93
|
3 |
+
lm_scale_0.7 4.94
|
4 |
+
lm_scale_0.5 4.95
|
5 |
+
lm_scale_0.4 4.99
|
6 |
+
lm_scale_0.8 4.99
|
7 |
+
lm_scale_0.9 5.03
|
8 |
+
lm_scale_0.3 5.04
|
9 |
+
lm_scale_1.0 5.08
|
10 |
+
lm_scale_0.2 5.1
|
11 |
+
lm_scale_1.1 5.16
|
12 |
+
lm_scale_0.1 5.2
|
13 |
+
lm_scale_1.2 5.25
|
14 |
+
lm_scale_1.3 5.3
|
15 |
+
lm_scale_1.4 5.38
|
16 |
+
lm_scale_1.5 5.45
|
17 |
+
lm_scale_1.6 5.52
|
18 |
+
lm_scale_1.7 5.58
|
19 |
+
lm_scale_1.8 5.63
|
20 |
+
lm_scale_1.9 5.7
|
21 |
+
lm_scale_2.0 5.74
|
decoding_results/nbest/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/nbest/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/nbest/log-decode-epoch-40-avg-16-use-averaged-model-2023-06-13-17-21-29
ADDED
@@ -0,0 +1,35 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
2023-06-13 17:21:29,434 INFO [ctc_decode.py:633] Decoding started
|
2 |
+
2023-06-13 17:21:29,434 INFO [ctc_decode.py:639] Device: cuda:0
|
3 |
+
2023-06-13 17:21:29,434 INFO [ctc_decode.py:640] {'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 50, 'reset_interval': 200, 'valid_interval': 3000, 'feature_dim': 80, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'c51a0b9684442a88ee37f3ce0af686a04b66855b', 'k2-git-date': 'Mon May 1 21:38:03 2023', 'lhotse-version': '1.12.0.dev+git.891bad1.clean', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'new-zipformer-add-ctc', 'icefall-git-sha1': '046b6cb6-clean', 'icefall-git-date': 'Fri Jun 2 15:51:49 2023', 'icefall-path': '/ceph-zw/workspace/zipformer/icefall_zipformer', 'k2-path': '/ceph-zw/workspace/k2/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-zw/workspace/share/lhotse/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-1-1220091118-57c4d55446-mlpzc', 'IP address': '10.177.22.19'}, 'frame_shift_ms': 10, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'use_double_scores': True, 'epoch': 40, 'iter': 0, 'avg': 16, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp-ctc-rnnt'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'context_size': 2, 'decoding_method': 'nbest', 'num_paths': 100, 'nbest_scale': 1.0, 'hlg_scale': 0.6, 'lm_dir': PosixPath('data/lm'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'decoder_dim': 512, 'joiner_dim': 512, 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'use_transducer': True, 'use_ctc': True, 'full_libri': True, 'mini_libri': False, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 300, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'drop_last': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp-ctc-rnnt/nbest'), 'suffix': 'epoch-40-avg-16-use-averaged-model'}
|
4 |
+
2023-06-13 17:21:29,725 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
|
5 |
+
2023-06-13 17:21:34,644 INFO [ctc_decode.py:718] About to create model
|
6 |
+
2023-06-13 17:21:35,244 INFO [ctc_decode.py:785] Calculating the averaged model over epoch range from 24 (excluded) to 40
|
7 |
+
2023-06-13 17:21:39,252 INFO [ctc_decode.py:802] Number of model parameters: 65805511
|
8 |
+
2023-06-13 17:21:39,252 INFO [asr_datamodule.py:465] About to get test-clean cuts
|
9 |
+
2023-06-13 17:21:39,255 INFO [asr_datamodule.py:472] About to get test-other cuts
|
10 |
+
2023-06-13 17:21:40,545 INFO [ctc_decode.py:550] batch 0/?, cuts processed until now is 21
|
11 |
+
2023-06-13 17:21:42,976 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([5.0315, 4.4660, 4.2767, 4.5975], device='cuda:0')
|
12 |
+
2023-06-13 17:21:55,627 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([4.4431, 2.6692, 3.6222, 2.0333], device='cuda:0')
|
13 |
+
2023-06-13 17:22:11,994 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([2.2852, 3.4389, 3.4427, 3.4871, 3.2347, 2.9755, 2.4241, 2.8685],
|
14 |
+
device='cuda:0')
|
15 |
+
2023-06-13 17:22:16,053 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([4.5884, 2.7881, 3.7633, 2.0764], device='cuda:0')
|
16 |
+
2023-06-13 17:22:20,226 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([3.9332, 4.9863, 4.5516, 4.1826], device='cuda:0')
|
17 |
+
2023-06-13 17:22:26,654 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([2.8336, 4.2831, 4.1287, 3.3007, 3.3161, 3.7073, 4.0190, 3.6861],
|
18 |
+
device='cuda:0')
|
19 |
+
2023-06-13 17:22:28,090 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
20 |
+
2023-06-13 17:22:28,169 INFO [utils.py:561] [test-clean-no_rescore-nbest-scale-1.0-100] %WER 2.46% [1294 / 52576, 181 ins, 93 del, 1020 sub ]
|
21 |
+
2023-06-13 17:22:28,344 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
22 |
+
2023-06-13 17:22:28,345 INFO [ctc_decode.py:587]
|
23 |
+
For test-clean, WER of different settings are:
|
24 |
+
no_rescore-nbest-scale-1.0-100 2.46 best for test-clean
|
25 |
+
|
26 |
+
2023-06-13 17:22:29,623 INFO [ctc_decode.py:550] batch 0/?, cuts processed until now is 26
|
27 |
+
2023-06-13 17:22:59,061 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([3.7139, 3.3612, 2.7698, 3.3808], device='cuda:0')
|
28 |
+
2023-06-13 17:23:25,555 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
29 |
+
2023-06-13 17:23:25,630 INFO [utils.py:561] [test-other-no_rescore-nbest-scale-1.0-100] %WER 5.11% [2674 / 52343, 288 ins, 256 del, 2130 sub ]
|
30 |
+
2023-06-13 17:23:25,808 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
31 |
+
2023-06-13 17:23:25,808 INFO [ctc_decode.py:587]
|
32 |
+
For test-other, WER of different settings are:
|
33 |
+
no_rescore-nbest-scale-1.0-100 5.11 best for test-other
|
34 |
+
|
35 |
+
2023-06-13 17:23:25,808 INFO [ctc_decode.py:835] Done!
|
decoding_results/nbest/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/nbest/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/nbest/wer-summary-test-clean-epoch-40-avg-16-use-averaged-model.txt
ADDED
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
1 |
+
settings WER
|
2 |
+
no_rescore-nbest-scale-1.0-100 2.46
|
decoding_results/nbest/wer-summary-test-other-epoch-40-avg-16-use-averaged-model.txt
ADDED
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
1 |
+
settings WER
|
2 |
+
no_rescore-nbest-scale-1.0-100 5.11
|
decoding_results/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/whole-lattice-rescoring/log-decode-epoch-40-avg-16-use-averaged-model-2023-06-13-17-33-26
ADDED
@@ -0,0 +1,251 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
2023-06-13 17:33:26,582 INFO [ctc_decode.py:633] Decoding started
|
2 |
+
2023-06-13 17:33:26,582 INFO [ctc_decode.py:639] Device: cuda:0
|
3 |
+
2023-06-13 17:33:26,582 INFO [ctc_decode.py:640] {'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 50, 'reset_interval': 200, 'valid_interval': 3000, 'feature_dim': 80, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'c51a0b9684442a88ee37f3ce0af686a04b66855b', 'k2-git-date': 'Mon May 1 21:38:03 2023', 'lhotse-version': '1.12.0.dev+git.891bad1.clean', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'new-zipformer-add-ctc', 'icefall-git-sha1': '046b6cb6-clean', 'icefall-git-date': 'Fri Jun 2 15:51:49 2023', 'icefall-path': '/ceph-zw/workspace/zipformer/icefall_zipformer', 'k2-path': '/ceph-zw/workspace/k2/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-zw/workspace/share/lhotse/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-1-1220091118-57c4d55446-mlpzc', 'IP address': '10.177.22.19'}, 'frame_shift_ms': 10, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'use_double_scores': True, 'epoch': 40, 'iter': 0, 'avg': 16, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp-ctc-rnnt'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'context_size': 2, 'decoding_method': 'whole-lattice-rescoring', 'num_paths': 100, 'nbest_scale': 1.0, 'hlg_scale': 0.6, 'lm_dir': PosixPath('data/my_lm'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'decoder_dim': 512, 'joiner_dim': 512, 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'use_transducer': True, 'use_ctc': True, 'full_libri': True, 'mini_libri': False, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 300, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'drop_last': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp-ctc-rnnt/whole-lattice-rescoring'), 'suffix': 'epoch-40-avg-16-use-averaged-model'}
|
4 |
+
2023-06-13 17:33:26,852 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
|
5 |
+
2023-06-13 17:33:31,836 INFO [ctc_decode.py:701] Loading pre-compiled G_4_gram.pt
|
6 |
+
2023-06-13 17:33:45,493 INFO [ctc_decode.py:718] About to create model
|
7 |
+
2023-06-13 17:33:46,067 INFO [ctc_decode.py:785] Calculating the averaged model over epoch range from 24 (excluded) to 40
|
8 |
+
2023-06-13 17:33:48,190 INFO [ctc_decode.py:802] Number of model parameters: 65805511
|
9 |
+
2023-06-13 17:33:48,190 INFO [asr_datamodule.py:465] About to get test-clean cuts
|
10 |
+
2023-06-13 17:33:48,193 INFO [asr_datamodule.py:472] About to get test-other cuts
|
11 |
+
2023-06-13 17:33:49,550 INFO [ctc_decode.py:550] batch 0/?, cuts processed until now is 21
|
12 |
+
2023-06-13 17:34:04,381 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([4.1588, 2.7219, 2.7286, 3.4695], device='cuda:0')
|
13 |
+
2023-06-13 17:34:43,425 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
14 |
+
2023-06-13 17:34:43,497 INFO [utils.py:561] [test-clean-lm_scale_0.1] %WER 2.51% [1318 / 52576, 228 ins, 66 del, 1024 sub ]
|
15 |
+
2023-06-13 17:34:43,669 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
16 |
+
2023-06-13 17:34:43,693 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
17 |
+
2023-06-13 17:34:43,761 INFO [utils.py:561] [test-clean-lm_scale_0.2] %WER 2.47% [1299 / 52576, 216 ins, 67 del, 1016 sub ]
|
18 |
+
2023-06-13 17:34:44,120 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
19 |
+
2023-06-13 17:34:44,143 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
20 |
+
2023-06-13 17:34:44,214 INFO [utils.py:561] [test-clean-lm_scale_0.3] %WER 2.42% [1272 / 52576, 203 ins, 71 del, 998 sub ]
|
21 |
+
2023-06-13 17:34:44,391 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
22 |
+
2023-06-13 17:34:44,413 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
23 |
+
2023-06-13 17:34:44,482 INFO [utils.py:561] [test-clean-lm_scale_0.4] %WER 2.39% [1254 / 52576, 192 ins, 74 del, 988 sub ]
|
24 |
+
2023-06-13 17:34:44,653 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
25 |
+
2023-06-13 17:34:44,675 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
26 |
+
2023-06-13 17:34:44,744 INFO [utils.py:561] [test-clean-lm_scale_0.5] %WER 2.37% [1245 / 52576, 184 ins, 81 del, 980 sub ]
|
27 |
+
2023-06-13 17:34:44,920 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
28 |
+
2023-06-13 17:34:44,943 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
29 |
+
2023-06-13 17:34:45,011 INFO [utils.py:561] [test-clean-lm_scale_0.6] %WER 2.38% [1250 / 52576, 178 ins, 87 del, 985 sub ]
|
30 |
+
2023-06-13 17:34:45,180 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
31 |
+
2023-06-13 17:34:45,202 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
32 |
+
2023-06-13 17:34:45,272 INFO [utils.py:561] [test-clean-lm_scale_0.7] %WER 2.39% [1259 / 52576, 171 ins, 98 del, 990 sub ]
|
33 |
+
2023-06-13 17:34:45,440 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
34 |
+
2023-06-13 17:34:45,462 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
35 |
+
2023-06-13 17:34:45,530 INFO [utils.py:561] [test-clean-lm_scale_0.8] %WER 2.43% [1277 / 52576, 164 ins, 113 del, 1000 sub ]
|
36 |
+
2023-06-13 17:34:45,698 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
37 |
+
2023-06-13 17:34:45,720 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
38 |
+
2023-06-13 17:34:45,790 INFO [utils.py:561] [test-clean-lm_scale_0.9] %WER 2.49% [1307 / 52576, 155 ins, 137 del, 1015 sub ]
|
39 |
+
2023-06-13 17:34:45,961 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
40 |
+
2023-06-13 17:34:45,983 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
41 |
+
2023-06-13 17:34:46,202 INFO [utils.py:561] [test-clean-lm_scale_1.0] %WER 2.57% [1350 / 52576, 151 ins, 162 del, 1037 sub ]
|
42 |
+
2023-06-13 17:34:46,373 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
43 |
+
2023-06-13 17:34:46,396 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
44 |
+
2023-06-13 17:34:46,464 INFO [utils.py:561] [test-clean-lm_scale_1.1] %WER 2.72% [1432 / 52576, 151 ins, 202 del, 1079 sub ]
|
45 |
+
2023-06-13 17:34:46,635 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
46 |
+
2023-06-13 17:34:46,657 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
47 |
+
2023-06-13 17:34:46,726 INFO [utils.py:561] [test-clean-lm_scale_1.2] %WER 2.84% [1493 / 52576, 148 ins, 236 del, 1109 sub ]
|
48 |
+
2023-06-13 17:34:46,898 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
49 |
+
2023-06-13 17:34:46,920 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
50 |
+
2023-06-13 17:34:46,988 INFO [utils.py:561] [test-clean-lm_scale_1.3] %WER 3.03% [1594 / 52576, 147 ins, 293 del, 1154 sub ]
|
51 |
+
2023-06-13 17:34:47,159 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
52 |
+
2023-06-13 17:34:47,181 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
53 |
+
2023-06-13 17:34:47,250 INFO [utils.py:561] [test-clean-lm_scale_1.4] %WER 3.22% [1695 / 52576, 144 ins, 360 del, 1191 sub ]
|
54 |
+
2023-06-13 17:34:47,421 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
55 |
+
2023-06-13 17:34:47,443 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
56 |
+
2023-06-13 17:34:47,513 INFO [utils.py:561] [test-clean-lm_scale_1.5] %WER 3.44% [1810 / 52576, 136 ins, 432 del, 1242 sub ]
|
57 |
+
2023-06-13 17:34:47,683 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
58 |
+
2023-06-13 17:34:47,705 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
59 |
+
2023-06-13 17:34:47,923 INFO [utils.py:561] [test-clean-lm_scale_1.6] %WER 3.69% [1939 / 52576, 136 ins, 510 del, 1293 sub ]
|
60 |
+
2023-06-13 17:34:48,094 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
61 |
+
2023-06-13 17:34:48,117 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
62 |
+
2023-06-13 17:34:48,185 INFO [utils.py:561] [test-clean-lm_scale_1.7] %WER 3.93% [2065 / 52576, 133 ins, 576 del, 1356 sub ]
|
63 |
+
2023-06-13 17:34:48,360 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
64 |
+
2023-06-13 17:34:48,382 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
65 |
+
2023-06-13 17:34:48,451 INFO [utils.py:561] [test-clean-lm_scale_1.8] %WER 4.11% [2163 / 52576, 131 ins, 629 del, 1403 sub ]
|
66 |
+
2023-06-13 17:34:48,621 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
67 |
+
2023-06-13 17:34:48,644 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
68 |
+
2023-06-13 17:34:48,711 INFO [utils.py:561] [test-clean-lm_scale_1.9] %WER 4.35% [2285 / 52576, 129 ins, 701 del, 1455 sub ]
|
69 |
+
2023-06-13 17:34:48,889 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
70 |
+
2023-06-13 17:34:48,911 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
71 |
+
2023-06-13 17:34:48,980 INFO [utils.py:561] [test-clean-lm_scale_2.0] %WER 4.58% [2409 / 52576, 130 ins, 767 del, 1512 sub ]
|
72 |
+
2023-06-13 17:34:49,152 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
|
73 |
+
2023-06-13 17:34:49,153 INFO [ctc_decode.py:587]
|
74 |
+
For test-clean, WER of different settings are:
|
75 |
+
lm_scale_0.5 2.37 best for test-clean
|
76 |
+
lm_scale_0.6 2.38
|
77 |
+
lm_scale_0.4 2.39
|
78 |
+
lm_scale_0.7 2.39
|
79 |
+
lm_scale_0.3 2.42
|
80 |
+
lm_scale_0.8 2.43
|
81 |
+
lm_scale_0.2 2.47
|
82 |
+
lm_scale_0.9 2.49
|
83 |
+
lm_scale_0.1 2.51
|
84 |
+
lm_scale_1.0 2.57
|
85 |
+
lm_scale_1.1 2.72
|
86 |
+
lm_scale_1.2 2.84
|
87 |
+
lm_scale_1.3 3.03
|
88 |
+
lm_scale_1.4 3.22
|
89 |
+
lm_scale_1.5 3.44
|
90 |
+
lm_scale_1.6 3.69
|
91 |
+
lm_scale_1.7 3.93
|
92 |
+
lm_scale_1.8 4.11
|
93 |
+
lm_scale_1.9 4.35
|
94 |
+
lm_scale_2.0 4.58
|
95 |
+
|
96 |
+
2023-06-13 17:34:50,631 INFO [ctc_decode.py:550] batch 0/?, cuts processed until now is 26
|
97 |
+
2023-06-13 17:35:10,848 INFO [decode.py:893] Caught exception:
|
98 |
+
CUDA out of memory. Tried to allocate 8.00 GiB (GPU 0; 31.75 GiB total capacity; 23.78 GiB already allocated; 181.69 MiB free; 30.20 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
|
99 |
+
Exception raised from malloc at ../c10/cuda/CUDACachingAllocator.cpp:536 (most recent call first):
|
100 |
+
frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x42 (0x7f34d9c0fd62 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10.so)
|
101 |
+
frame #1: <unknown function> + 0x25358 (0x7f34d9e7b358 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10_cuda.so)
|
102 |
+
frame #2: <unknown function> + 0x25d72 (0x7f34d9e7bd72 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10_cuda.so)
|
103 |
+
frame #3: <unknown function> + 0x261a2 (0x7f34d9e7c1a2 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10_cuda.so)
|
104 |
+
frame #4: k2::PytorchCudaContext::Allocate(unsigned long, void**) + 0x32 (0x7f34328d10d2 in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
|
105 |
+
frame #5: k2::NewRegion(std::shared_ptr<k2::Context>, unsigned long) + 0x112 (0x7f34325f6392 in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
|
106 |
+
frame #6: k2::Hash::Hash(std::shared_ptr<k2::Context>, int, int, int) + 0x2f7 (0x7f34326cb3a7 in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
|
107 |
+
frame #7: k2::Hash::Resize(int, int, int, bool) + 0x1b4 (0x7f34326c0eb4 in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
|
108 |
+
frame #8: k2::DeviceIntersector::ForwardSortedA() + 0x53e (0x7f34326f5fae in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
|
109 |
+
frame #9: k2::IntersectDevice(k2::Ragged<k2::Arc>&, int, k2::Ragged<k2::Arc>&, int, k2::Array1<int> const&, k2::Array1<int>*, k2::Array1<int>*, bool) + 0x4cd (0x7f34326d81fd in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
|
110 |
+
frame #10: <unknown function> + 0x75936 (0x7f3433715936 in /ceph-zw/workspace/k2/k2/build_release/lib/_k2.cpython-38-x86_64-linux-gnu.so)
|
111 |
+
frame #11: <unknown function> + 0x377ec (0x7f34336d77ec in /ceph-zw/workspace/k2/k2/build_release/lib/_k2.cpython-38-x86_64-linux-gnu.so)
|
112 |
+
<omitting python frames>
|
113 |
+
frame #34: python3() [0x662c2e]
|
114 |
+
frame #39: __libc_start_main + 0xe7 (0x7f353b2a7bf7 in /lib/x86_64-linux-gnu/libc.so.6)
|
115 |
+
|
116 |
+
|
117 |
+
2023-06-13 17:35:10,848 INFO [decode.py:897] num_arcs before pruning: 262949
|
118 |
+
2023-06-13 17:35:10,848 INFO [decode.py:898] This OOM is not an error. You can ignore it. If your model does not converge well, or --max-duration is too large, or the input sound file is difficult to decode, you will meet this exception.
|
119 |
+
2023-06-13 17:35:10,857 INFO [decode.py:909] num_arcs after pruning: 45772
|
120 |
+
2023-06-13 17:35:36,889 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([3.4643, 2.3418, 2.5677, 2.0390], device='cuda:0')
|
121 |
+
2023-06-13 17:35:37,983 INFO [decode.py:893] Caught exception:
|
122 |
+
CUDA out of memory. Tried to allocate 2.00 GiB (GPU 0; 31.75 GiB total capacity; 29.74 GiB already allocated; 27.69 MiB free; 30.35 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
|
123 |
+
Exception raised from malloc at ../c10/cuda/CUDACachingAllocator.cpp:536 (most recent call first):
|
124 |
+
frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x42 (0x7f34d9c0fd62 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10.so)
|
125 |
+
frame #1: <unknown function> + 0x25358 (0x7f34d9e7b358 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10_cuda.so)
|
126 |
+
frame #2: <unknown function> + 0x25d72 (0x7f34d9e7bd72 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10_cuda.so)
|
127 |
+
frame #3: <unknown function> + 0x261a2 (0x7f34d9e7c1a2 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10_cuda.so)
|
128 |
+
frame #4: k2::PytorchCudaContext::Allocate(unsigned long, void**) + 0x32 (0x7f34328d10d2 in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
|
129 |
+
frame #5: k2::Region::Extend(unsigned long) + 0x7a (0x7f34326dcaea in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
|
130 |
+
frame #6: k2::Array1<k2::intersect_internal::StateInfo>::Resize(int, bool) + 0xb5 (0x7f34326dd435 in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
|
131 |
+
frame #7: void k2::DeviceIntersector::ForwardSortedAOneIter<k2::Hash::PackedAccessor>(int, k2::Array1<int> const&, k2::Array1<int> const&, k2::Array1<int> const&, k2::Array1<int> const&, int) + 0x9da (0x7f34326f4d8a in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
|
132 |
+
frame #8: k2::DeviceIntersector::ForwardSortedA() + 0xadb (0x7f34326f654b in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
|
133 |
+
frame #9: k2::IntersectDevice(k2::Ragged<k2::Arc>&, int, k2::Ragged<k2::Arc>&, int, k2::Array1<int> const&, k2::Array1<int>*, k2::Array1<int>*, bool) + 0x4cd (0x7f34326d81fd in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
|
134 |
+
frame #10: <unknown function> + 0x75936 (0x7f3433715936 in /ceph-zw/workspace/k2/k2/build_release/lib/_k2.cpython-38-x86_64-linux-gnu.so)
|
135 |
+
frame #11: <unknown function> + 0x377ec (0x7f34336d77ec in /ceph-zw/workspace/k2/k2/build_release/lib/_k2.cpython-38-x86_64-linux-gnu.so)
|
136 |
+
<omitting python frames>
|
137 |
+
frame #34: python3() [0x662c2e]
|
138 |
+
frame #39: __libc_start_main + 0xe7 (0x7f353b2a7bf7 in /lib/x86_64-linux-gnu/libc.so.6)
|
139 |
+
|
140 |
+
|
141 |
+
2023-06-13 17:35:37,983 INFO [decode.py:897] num_arcs before pruning: 305119
|
142 |
+
2023-06-13 17:35:37,983 INFO [decode.py:898] This OOM is not an error. You can ignore it. If your model does not converge well, or --max-duration is too large, or the input sound file is difficult to decode, you will meet this exception.
|
143 |
+
2023-06-13 17:35:38,002 INFO [decode.py:909] num_arcs after pruning: 45058
|
144 |
+
2023-06-13 17:35:49,399 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([4.0697, 2.1705, 3.1589, 1.7420], device='cuda:0')
|
145 |
+
2023-06-13 17:35:51,043 INFO [decode.py:893] Caught exception:
|
146 |
+
CUDA out of memory. Tried to allocate 2.00 GiB (GPU 0; 31.75 GiB total capacity; 28.59 GiB already allocated; 813.69 MiB free; 29.58 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
|
147 |
+
Exception raised from malloc at ../c10/cuda/CUDACachingAllocator.cpp:536 (most recent call first):
|
148 |
+
frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x42 (0x7f34d9c0fd62 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10.so)
|
149 |
+
frame #1: <unknown function> + 0x25358 (0x7f34d9e7b358 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10_cuda.so)
|
150 |
+
frame #2: <unknown function> + 0x25d72 (0x7f34d9e7bd72 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10_cuda.so)
|
151 |
+
frame #3: <unknown function> + 0x261a2 (0x7f34d9e7c1a2 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10_cuda.so)
|
152 |
+
frame #4: k2::PytorchCudaContext::Allocate(unsigned long, void**) + 0x32 (0x7f34328d10d2 in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
|
153 |
+
frame #5: k2::Region::Extend(unsigned long) + 0x7a (0x7f34326dcaea in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
|
154 |
+
frame #6: k2::Array1<k2::intersect_internal::StateInfo>::Resize(int, bool) + 0xb5 (0x7f34326dd435 in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
|
155 |
+
frame #7: void k2::DeviceIntersector::ForwardSortedAOneIter<k2::Hash::PackedAccessor>(int, k2::Array1<int> const&, k2::Array1<int> const&, k2::Array1<int> const&, k2::Array1<int> const&, int) + 0x9da (0x7f34326f4d8a in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
|
156 |
+
frame #8: k2::DeviceIntersector::ForwardSortedA() + 0xadb (0x7f34326f654b in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
|
157 |
+
frame #9: k2::IntersectDevice(k2::Ragged<k2::Arc>&, int, k2::Ragged<k2::Arc>&, int, k2::Array1<int> const&, k2::Array1<int>*, k2::Array1<int>*, bool) + 0x4cd (0x7f34326d81fd in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
|
158 |
+
frame #10: <unknown function> + 0x75936 (0x7f3433715936 in /ceph-zw/workspace/k2/k2/build_release/lib/_k2.cpython-38-x86_64-linux-gnu.so)
|
159 |
+
frame #11: <unknown function> + 0x377ec (0x7f34336d77ec in /ceph-zw/workspace/k2/k2/build_release/lib/_k2.cpython-38-x86_64-linux-gnu.so)
|
160 |
+
<omitting python frames>
|
161 |
+
frame #34: python3() [0x662c2e]
|
162 |
+
frame #39: __libc_start_main + 0xe7 (0x7f353b2a7bf7 in /lib/x86_64-linux-gnu/libc.so.6)
|
163 |
+
|
164 |
+
|
165 |
+
2023-06-13 17:35:51,043 INFO [decode.py:897] num_arcs before pruning: 347421
|
166 |
+
2023-06-13 17:35:51,043 INFO [decode.py:898] This OOM is not an error. You can ignore it. If your model does not converge well, or --max-duration is too large, or the input sound file is difficult to decode, you will meet this exception.
|
167 |
+
2023-06-13 17:35:51,064 INFO [decode.py:909] num_arcs after pruning: 22726
|
168 |
+
2023-06-13 17:35:52,877 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
169 |
+
2023-06-13 17:35:52,951 INFO [utils.py:561] [test-other-lm_scale_0.1] %WER 5.29% [2768 / 52343, 385 ins, 176 del, 2207 sub ]
|
170 |
+
2023-06-13 17:35:53,126 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
171 |
+
2023-06-13 17:35:53,150 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
172 |
+
2023-06-13 17:35:53,226 INFO [utils.py:561] [test-other-lm_scale_0.2] %WER 5.13% [2686 / 52343, 365 ins, 176 del, 2145 sub ]
|
173 |
+
2023-06-13 17:35:53,401 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
174 |
+
2023-06-13 17:35:53,424 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
175 |
+
2023-06-13 17:35:53,493 INFO [utils.py:561] [test-other-lm_scale_0.3] %WER 5.04% [2639 / 52343, 346 ins, 185 del, 2108 sub ]
|
176 |
+
2023-06-13 17:35:53,666 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
177 |
+
2023-06-13 17:35:53,688 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
178 |
+
2023-06-13 17:35:53,758 INFO [utils.py:561] [test-other-lm_scale_0.4] %WER 4.98% [2608 / 52343, 325 ins, 196 del, 2087 sub ]
|
179 |
+
2023-06-13 17:35:53,933 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
180 |
+
2023-06-13 17:35:53,955 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
181 |
+
2023-06-13 17:35:54,025 INFO [utils.py:561] [test-other-lm_scale_0.5] %WER 4.93% [2581 / 52343, 315 ins, 212 del, 2054 sub ]
|
182 |
+
2023-06-13 17:35:54,198 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
183 |
+
2023-06-13 17:35:54,221 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
184 |
+
2023-06-13 17:35:54,291 INFO [utils.py:561] [test-other-lm_scale_0.6] %WER 4.91% [2568 / 52343, 290 ins, 231 del, 2047 sub ]
|
185 |
+
2023-06-13 17:35:54,464 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
186 |
+
2023-06-13 17:35:54,487 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
187 |
+
2023-06-13 17:35:54,557 INFO [utils.py:561] [test-other-lm_scale_0.7] %WER 4.88% [2556 / 52343, 272 ins, 256 del, 2028 sub ]
|
188 |
+
2023-06-13 17:35:54,728 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
189 |
+
2023-06-13 17:35:54,751 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
190 |
+
2023-06-13 17:35:54,821 INFO [utils.py:561] [test-other-lm_scale_0.8] %WER 4.97% [2600 / 52343, 259 ins, 303 del, 2038 sub ]
|
191 |
+
2023-06-13 17:35:54,995 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
192 |
+
2023-06-13 17:35:55,017 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
193 |
+
2023-06-13 17:35:55,086 INFO [utils.py:561] [test-other-lm_scale_0.9] %WER 5.07% [2652 / 52343, 251 ins, 347 del, 2054 sub ]
|
194 |
+
2023-06-13 17:35:55,260 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
195 |
+
2023-06-13 17:35:55,283 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
196 |
+
2023-06-13 17:35:55,353 INFO [utils.py:561] [test-other-lm_scale_1.0] %WER 5.16% [2703 / 52343, 239 ins, 400 del, 2064 sub ]
|
197 |
+
2023-06-13 17:35:55,679 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
198 |
+
2023-06-13 17:35:55,702 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
199 |
+
2023-06-13 17:35:55,772 INFO [utils.py:561] [test-other-lm_scale_1.1] %WER 5.36% [2805 / 52343, 230 ins, 461 del, 2114 sub ]
|
200 |
+
2023-06-13 17:35:55,945 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
201 |
+
2023-06-13 17:35:55,968 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
202 |
+
2023-06-13 17:35:56,039 INFO [utils.py:561] [test-other-lm_scale_1.2] %WER 5.60% [2933 / 52343, 220 ins, 570 del, 2143 sub ]
|
203 |
+
2023-06-13 17:35:56,214 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
204 |
+
2023-06-13 17:35:56,237 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
205 |
+
2023-06-13 17:35:56,313 INFO [utils.py:561] [test-other-lm_scale_1.3] %WER 5.88% [3079 / 52343, 214 ins, 691 del, 2174 sub ]
|
206 |
+
2023-06-13 17:35:56,489 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
207 |
+
2023-06-13 17:35:56,512 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
208 |
+
2023-06-13 17:35:56,584 INFO [utils.py:561] [test-other-lm_scale_1.4] %WER 6.16% [3226 / 52343, 214 ins, 780 del, 2232 sub ]
|
209 |
+
2023-06-13 17:35:56,759 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
210 |
+
2023-06-13 17:35:56,782 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
211 |
+
2023-06-13 17:35:56,854 INFO [utils.py:561] [test-other-lm_scale_1.5] %WER 6.52% [3411 / 52343, 215 ins, 898 del, 2298 sub ]
|
212 |
+
2023-06-13 17:35:57,029 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
213 |
+
2023-06-13 17:35:57,052 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
214 |
+
2023-06-13 17:35:57,120 INFO [utils.py:561] [test-other-lm_scale_1.6] %WER 6.80% [3559 / 52343, 206 ins, 1009 del, 2344 sub ]
|
215 |
+
2023-06-13 17:35:57,294 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
216 |
+
2023-06-13 17:35:57,317 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
217 |
+
2023-06-13 17:35:57,386 INFO [utils.py:561] [test-other-lm_scale_1.7] %WER 7.13% [3731 / 52343, 204 ins, 1139 del, 2388 sub ]
|
218 |
+
2023-06-13 17:35:57,712 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
219 |
+
2023-06-13 17:35:57,734 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
220 |
+
2023-06-13 17:35:57,805 INFO [utils.py:561] [test-other-lm_scale_1.8] %WER 7.42% [3886 / 52343, 198 ins, 1248 del, 2440 sub ]
|
221 |
+
2023-06-13 17:35:57,980 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
222 |
+
2023-06-13 17:35:58,003 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
223 |
+
2023-06-13 17:35:58,072 INFO [utils.py:561] [test-other-lm_scale_1.9] %WER 7.75% [4059 / 52343, 194 ins, 1359 del, 2506 sub ]
|
224 |
+
2023-06-13 17:35:58,245 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
225 |
+
2023-06-13 17:35:58,268 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
226 |
+
2023-06-13 17:35:58,338 INFO [utils.py:561] [test-other-lm_scale_2.0] %WER 8.04% [4206 / 52343, 190 ins, 1459 del, 2557 sub ]
|
227 |
+
2023-06-13 17:35:58,511 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
|
228 |
+
2023-06-13 17:35:58,512 INFO [ctc_decode.py:587]
|
229 |
+
For test-other, WER of different settings are:
|
230 |
+
lm_scale_0.7 4.88 best for test-other
|
231 |
+
lm_scale_0.6 4.91
|
232 |
+
lm_scale_0.5 4.93
|
233 |
+
lm_scale_0.8 4.97
|
234 |
+
lm_scale_0.4 4.98
|
235 |
+
lm_scale_0.3 5.04
|
236 |
+
lm_scale_0.9 5.07
|
237 |
+
lm_scale_0.2 5.13
|
238 |
+
lm_scale_1.0 5.16
|
239 |
+
lm_scale_0.1 5.29
|
240 |
+
lm_scale_1.1 5.36
|
241 |
+
lm_scale_1.2 5.6
|
242 |
+
lm_scale_1.3 5.88
|
243 |
+
lm_scale_1.4 6.16
|
244 |
+
lm_scale_1.5 6.52
|
245 |
+
lm_scale_1.6 6.8
|
246 |
+
lm_scale_1.7 7.13
|
247 |
+
lm_scale_1.8 7.42
|
248 |
+
lm_scale_1.9 7.75
|
249 |
+
lm_scale_2.0 8.04
|
250 |
+
|
251 |
+
2023-06-13 17:35:58,512 INFO [ctc_decode.py:835] Done!
|
decoding_results/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
decoding_results/whole-lattice-rescoring/wer-summary-test-clean-epoch-40-avg-16-use-averaged-model.txt
ADDED
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
settings WER
|
2 |
+
lm_scale_0.5 2.37
|
3 |
+
lm_scale_0.6 2.38
|
4 |
+
lm_scale_0.4 2.39
|
5 |
+
lm_scale_0.7 2.39
|
6 |
+
lm_scale_0.3 2.42
|
7 |
+
lm_scale_0.8 2.43
|
8 |
+
lm_scale_0.2 2.47
|
9 |
+
lm_scale_0.9 2.49
|
10 |
+
lm_scale_0.1 2.51
|
11 |
+
lm_scale_1.0 2.57
|
12 |
+
lm_scale_1.1 2.72
|
13 |
+
lm_scale_1.2 2.84
|
14 |
+
lm_scale_1.3 3.03
|
15 |
+
lm_scale_1.4 3.22
|
16 |
+
lm_scale_1.5 3.44
|
17 |
+
lm_scale_1.6 3.69
|
18 |
+
lm_scale_1.7 3.93
|
19 |
+
lm_scale_1.8 4.11
|
20 |
+
lm_scale_1.9 4.35
|
21 |
+
lm_scale_2.0 4.58
|
decoding_results/whole-lattice-rescoring/wer-summary-test-other-epoch-40-avg-16-use-averaged-model.txt
ADDED
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
settings WER
|
2 |
+
lm_scale_0.7 4.88
|
3 |
+
lm_scale_0.6 4.91
|
4 |
+
lm_scale_0.5 4.93
|
5 |
+
lm_scale_0.8 4.97
|
6 |
+
lm_scale_0.4 4.98
|
7 |
+
lm_scale_0.3 5.04
|
8 |
+
lm_scale_0.9 5.07
|
9 |
+
lm_scale_0.2 5.13
|
10 |
+
lm_scale_1.0 5.16
|
11 |
+
lm_scale_0.1 5.29
|
12 |
+
lm_scale_1.1 5.36
|
13 |
+
lm_scale_1.2 5.6
|
14 |
+
lm_scale_1.3 5.88
|
15 |
+
lm_scale_1.4 6.16
|
16 |
+
lm_scale_1.5 6.52
|
17 |
+
lm_scale_1.6 6.8
|
18 |
+
lm_scale_1.7 7.13
|
19 |
+
lm_scale_1.8 7.42
|
20 |
+
lm_scale_1.9 7.75
|
21 |
+
lm_scale_2.0 8.04
|
exp/decode.sh
ADDED
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
export CUDA_VISIBLE_DEVICES="0"
|
2 |
+
for m in ctc-decoding 1best nbest nbest-rescoring whole-lattice-rescoring; do
|
3 |
+
./zipformer/ctc_decode.py \
|
4 |
+
--epoch 40 \
|
5 |
+
--avg 16 \
|
6 |
+
--exp-dir zipformer/exp-ctc-rnnt \
|
7 |
+
--use-transducer 1 \
|
8 |
+
--use-ctc 1 \
|
9 |
+
--max-duration 300 \
|
10 |
+
--causal 0 \
|
11 |
+
--num-paths 100 \
|
12 |
+
--nbest-scale 1.0 \
|
13 |
+
--hlg-scale 0.6 \
|
14 |
+
--decoding-method $m
|
15 |
+
done
|
16 |
+
|
exp/epoch-40.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:fee348148cdb5910ad116ad0fcfa6469f5976e0b1552d9b1899a3b3093c18d39
|
3 |
+
size 1053871436
|
exp/export.sh
ADDED
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
./zipformer/export.py \
|
2 |
+
--exp-dir ./zipformer/exp-ctc-rnnt \
|
3 |
+
--use-transducer 1 \
|
4 |
+
--use-ctc 1 \
|
5 |
+
--bpe-model data/lang_bpe_500/bpe.model \
|
6 |
+
--epoch 40 \
|
7 |
+
--avg 16 \
|
8 |
+
--jit 1
|
9 |
+
|
10 |
+
|
exp/jit_script.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a877a5d18386311fe069dec68bb7e0ca956129b90f8a4ce4e956dad1ac98e154
|
3 |
+
size 265955146
|
exp/log/log-train-2023-06-01-20-30-01-0
ADDED
The diff for this file is too large to render.
See raw diff
|
|