Hugging Face's logo

Spaces:
AudioLLMs
/
AudioBench-Leaderboard-Extend
Running

App Files Community
zhuohan-7 commited on
Commit
6c5c54e
·
verified ·
1 Parent(s): e400118

Upload folder using huggingface_hub

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. .gitattributes +1 -0
  2. examples/AC/AudioCaps-Test/data-00000-of-00001.arrow +2 -2
  3. examples/AC/AudioCaps-Test/dataset_info.json +18 -0
  4. examples/AC/AudioCaps-Test/sample_0.wav +0 -0
  5. examples/AC/AudioCaps-Test/sample_1.wav +0 -0
  6. examples/AC/AudioCaps-Test/sample_2.wav +0 -0
  7. examples/AC/AudioCaps-Test/state.json +2 -1
  8. examples/AC/WavCaps-Test/data-00000-of-00001.arrow +2 -2
  9. examples/AC/WavCaps-Test/dataset_info.json +18 -0
  10. examples/AC/WavCaps-Test/sample_0.wav +0 -0
  11. examples/AC/WavCaps-Test/sample_1.wav +0 -0
  12. examples/AC/WavCaps-Test/sample_2.wav +0 -0
  13. examples/AC/WavCaps-Test/state.json +2 -1
  14. examples/ASR/Common-Voice-15-En-Test/data-00000-of-00001.arrow +2 -2
  15. examples/ASR/Common-Voice-15-En-Test/dataset_info.json +21 -3
  16. examples/ASR/Common-Voice-15-En-Test/sample_0.wav +0 -0
  17. examples/ASR/Common-Voice-15-En-Test/sample_1.wav +0 -0
  18. examples/ASR/Common-Voice-15-En-Test/sample_2.wav +0 -0
  19. examples/ASR/Common-Voice-15-En-Test/state.json +2 -1
  20. examples/ASR/Earnings21-Test/data-00000-of-00001.arrow +2 -2
  21. examples/ASR/Earnings21-Test/dataset_info.json +18 -0
  22. examples/ASR/Earnings21-Test/sample_0.wav +2 -2
  23. examples/ASR/Earnings21-Test/sample_1.wav +2 -2
  24. examples/ASR/Earnings21-Test/sample_2.wav +2 -2
  25. examples/ASR/Earnings21-Test/state.json +2 -1
  26. examples/ASR/Earnings22-Test/data-00000-of-00001.arrow +2 -2
  27. examples/ASR/Earnings22-Test/dataset_info.json +18 -0
  28. examples/ASR/Earnings22-Test/sample_0.wav +2 -2
  29. examples/ASR/Earnings22-Test/sample_1.wav +2 -2
  30. examples/ASR/Earnings22-Test/sample_2.wav +2 -2
  31. examples/ASR/Earnings22-Test/state.json +2 -1
  32. examples/ASR/GigaSpeech-Test/data-00000-of-00001.arrow +2 -2
  33. examples/ASR/GigaSpeech-Test/dataset_info.json +18 -0
  34. examples/ASR/GigaSpeech-Test/sample_0.wav +0 -0
  35. examples/ASR/GigaSpeech-Test/sample_1.wav +0 -0
  36. examples/ASR/GigaSpeech-Test/sample_2.wav +0 -0
  37. examples/ASR/GigaSpeech-Test/state.json +2 -1
  38. examples/ASR/IMDA-Part1-ASR-Test/data-00000-of-00001.arrow +2 -2
  39. examples/ASR/IMDA-Part1-ASR-Test/sample_0.wav +0 -0
  40. examples/ASR/IMDA-Part1-ASR-Test/sample_1.wav +0 -0
  41. examples/ASR/IMDA-Part1-ASR-Test/sample_2.wav +0 -0
  42. examples/ASR/IMDA-Part1-ASR-Test/state.json +1 -1
  43. examples/ASR/IMDA-Part2-ASR-Test/data-00000-of-00001.arrow +2 -2
  44. examples/ASR/IMDA-Part2-ASR-Test/sample_0.wav +0 -0
  45. examples/ASR/IMDA-Part2-ASR-Test/sample_1.wav +0 -0
  46. examples/ASR/IMDA-Part2-ASR-Test/sample_2.wav +0 -0
  47. examples/ASR/IMDA-Part2-ASR-Test/state.json +1 -1
  48. examples/ASR/LibriSpeech-Test-Clean/data-00000-of-00001.arrow +2 -2
  49. examples/ASR/LibriSpeech-Test-Clean/dataset_info.json +18 -0
  50. examples/ASR/LibriSpeech-Test-Clean/sample_0.wav +0 -0
.gitattributes CHANGED
@@ -51,3 +51,4 @@ examples/SQA/SLUE-P2-SQA5-Test/sample_2.wav filter=lfs diff=lfs merge=lfs -text
51
  examples/SQA/Spoken-Squad-v1/sample_0.wav filter=lfs diff=lfs merge=lfs -text
52
  examples/SQA/Spoken-Squad-v1/sample_1.wav filter=lfs diff=lfs merge=lfs -text
53
  examples/SQA/Spoken-Squad-v1/sample_2.wav filter=lfs diff=lfs merge=lfs -text
 
 
51
  examples/SQA/Spoken-Squad-v1/sample_0.wav filter=lfs diff=lfs merge=lfs -text
52
  examples/SQA/Spoken-Squad-v1/sample_1.wav filter=lfs diff=lfs merge=lfs -text
53
  examples/SQA/Spoken-Squad-v1/sample_2.wav filter=lfs diff=lfs merge=lfs -text
54
+ examples/SQA/DREAM-TTS-MCQ-Test/sample_2.wav filter=lfs diff=lfs merge=lfs -text
examples/AC/AudioCaps-Test/data-00000-of-00001.arrow CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ff53d7bcff873fea97f02e971836f24adfbfba9ebdcfc7554423bf65cb110e1d
3
- size 970312
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:de5d651b3d63e461daae712fdcb437b28c163eda771ee610433cacc2c359c950
3
+ size 974008
examples/AC/AudioCaps-Test/dataset_info.json CHANGED
@@ -118,6 +118,24 @@
118
  "_type": "Value"
119
  }
120
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
121
  "qwen_audio_chat": {
122
  "answer": {
123
  "dtype": "string",
 
118
  "_type": "Value"
119
  }
120
  },
121
+ "mowe_audio": {
122
+ "answer": {
123
+ "dtype": "string",
124
+ "_type": "Value"
125
+ },
126
+ "model_prediction": {
127
+ "dtype": "string",
128
+ "_type": "Value"
129
+ },
130
+ "task_type": {
131
+ "dtype": "string",
132
+ "_type": "Value"
133
+ },
134
+ "text": {
135
+ "dtype": "string",
136
+ "_type": "Value"
137
+ }
138
+ },
139
  "qwen_audio_chat": {
140
  "answer": {
141
  "dtype": "string",
examples/AC/AudioCaps-Test/sample_0.wav CHANGED
Binary files a/examples/AC/AudioCaps-Test/sample_0.wav and b/examples/AC/AudioCaps-Test/sample_0.wav differ
 
examples/AC/AudioCaps-Test/sample_1.wav CHANGED
Binary files a/examples/AC/AudioCaps-Test/sample_1.wav and b/examples/AC/AudioCaps-Test/sample_1.wav differ
 
examples/AC/AudioCaps-Test/sample_2.wav CHANGED
Binary files a/examples/AC/AudioCaps-Test/sample_2.wav and b/examples/AC/AudioCaps-Test/sample_2.wav differ
 
examples/AC/AudioCaps-Test/state.json CHANGED
@@ -4,7 +4,7 @@
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
- "_fingerprint": "0e301916c3676d35",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
@@ -14,6 +14,7 @@
14
  "wavllm_fairseq",
15
  "Qwen2-Audio-7B-Instruct",
16
  "whisper_large_v3_with_llama_3_8b_instruct",
 
17
  "qwen_audio_chat"
18
  ],
19
  "_format_kwargs": {},
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "e654a4081bc1365b",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
 
14
  "wavllm_fairseq",
15
  "Qwen2-Audio-7B-Instruct",
16
  "whisper_large_v3_with_llama_3_8b_instruct",
17
+ "mowe_audio",
18
  "qwen_audio_chat"
19
  ],
20
  "_format_kwargs": {},
examples/AC/WavCaps-Test/data-00000-of-00001.arrow CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c824145183da91da498c0338860b6daeaed94d7c619ec9ae22a041918c61c902
3
- size 985360
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ca86f9ee00453c956db08be9d77e62a74a9accb5128c3dec0cb91a564a653ed3
3
+ size 529952
examples/AC/WavCaps-Test/dataset_info.json CHANGED
@@ -114,6 +114,24 @@
114
  "_type": "Value"
115
  }
116
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
117
  "qwen_audio_chat": {
118
  "answer": {
119
  "dtype": "string",
 
114
  "_type": "Value"
115
  }
116
  },
117
+ "mowe_audio": {
118
+ "answer": {
119
+ "dtype": "string",
120
+ "_type": "Value"
121
+ },
122
+ "model_prediction": {
123
+ "dtype": "string",
124
+ "_type": "Value"
125
+ },
126
+ "task_type": {
127
+ "dtype": "string",
128
+ "_type": "Value"
129
+ },
130
+ "text": {
131
+ "dtype": "string",
132
+ "_type": "Value"
133
+ }
134
+ },
135
  "qwen_audio_chat": {
136
  "answer": {
137
  "dtype": "string",
examples/AC/WavCaps-Test/sample_0.wav CHANGED
Binary files a/examples/AC/WavCaps-Test/sample_0.wav and b/examples/AC/WavCaps-Test/sample_0.wav differ
 
examples/AC/WavCaps-Test/sample_1.wav CHANGED
Binary files a/examples/AC/WavCaps-Test/sample_1.wav and b/examples/AC/WavCaps-Test/sample_1.wav differ
 
examples/AC/WavCaps-Test/sample_2.wav CHANGED
Binary files a/examples/AC/WavCaps-Test/sample_2.wav and b/examples/AC/WavCaps-Test/sample_2.wav differ
 
examples/AC/WavCaps-Test/state.json CHANGED
@@ -4,7 +4,7 @@
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
- "_fingerprint": "1e570096603c2a32",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
@@ -14,6 +14,7 @@
14
  "wavllm_fairseq",
15
  "Qwen2-Audio-7B-Instruct",
16
  "whisper_large_v3_with_llama_3_8b_instruct",
 
17
  "qwen_audio_chat"
18
  ],
19
  "_format_kwargs": {},
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "ce408e4cfa3eec8a",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
 
14
  "wavllm_fairseq",
15
  "Qwen2-Audio-7B-Instruct",
16
  "whisper_large_v3_with_llama_3_8b_instruct",
17
+ "mowe_audio",
18
  "qwen_audio_chat"
19
  ],
20
  "_format_kwargs": {},
examples/ASR/Common-Voice-15-En-Test/data-00000-of-00001.arrow CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:44587f333a31075993f3cc8832f8946fd59605246769dd8bf86887dcf3bd889c
3
- size 543912
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9b59c848db4f23898202a290a248174e7f05709c7cb32dbcf003afd40bf805ba
3
+ size 531272
examples/ASR/Common-Voice-15-En-Test/dataset_info.json CHANGED
@@ -34,11 +34,11 @@
34
  },
35
  "other_attributes": {
36
  "accents": {
37
- "dtype": "null",
38
  "_type": "Value"
39
  },
40
  "age": {
41
- "dtype": "null",
42
  "_type": "Value"
43
  },
44
  "client_id": {
@@ -50,7 +50,7 @@
50
  "_type": "Value"
51
  },
52
  "gender": {
53
- "dtype": "null",
54
  "_type": "Value"
55
  },
56
  "language": {
@@ -146,6 +146,24 @@
146
  "_type": "Value"
147
  }
148
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
149
  "qwen_audio_chat": {
150
  "answer": {
151
  "dtype": "string",
 
34
  },
35
  "other_attributes": {
36
  "accents": {
37
+ "dtype": "string",
38
  "_type": "Value"
39
  },
40
  "age": {
41
+ "dtype": "string",
42
  "_type": "Value"
43
  },
44
  "client_id": {
 
50
  "_type": "Value"
51
  },
52
  "gender": {
53
+ "dtype": "string",
54
  "_type": "Value"
55
  },
56
  "language": {
 
146
  "_type": "Value"
147
  }
148
  },
149
+ "mowe_audio": {
150
+ "answer": {
151
+ "dtype": "string",
152
+ "_type": "Value"
153
+ },
154
+ "model_prediction": {
155
+ "dtype": "string",
156
+ "_type": "Value"
157
+ },
158
+ "task_type": {
159
+ "dtype": "string",
160
+ "_type": "Value"
161
+ },
162
+ "text": {
163
+ "dtype": "string",
164
+ "_type": "Value"
165
+ }
166
+ },
167
  "qwen_audio_chat": {
168
  "answer": {
169
  "dtype": "string",
examples/ASR/Common-Voice-15-En-Test/sample_0.wav CHANGED
Binary files a/examples/ASR/Common-Voice-15-En-Test/sample_0.wav and b/examples/ASR/Common-Voice-15-En-Test/sample_0.wav differ
 
examples/ASR/Common-Voice-15-En-Test/sample_1.wav CHANGED
Binary files a/examples/ASR/Common-Voice-15-En-Test/sample_1.wav and b/examples/ASR/Common-Voice-15-En-Test/sample_1.wav differ
 
examples/ASR/Common-Voice-15-En-Test/sample_2.wav CHANGED
Binary files a/examples/ASR/Common-Voice-15-En-Test/sample_2.wav and b/examples/ASR/Common-Voice-15-En-Test/sample_2.wav differ
 
examples/ASR/Common-Voice-15-En-Test/state.json CHANGED
@@ -4,7 +4,7 @@
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
- "_fingerprint": "30218d56801da2e8",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
@@ -14,6 +14,7 @@
14
  "wavllm_fairseq",
15
  "Qwen2-Audio-7B-Instruct",
16
  "whisper_large_v3_with_llama_3_8b_instruct",
 
17
  "qwen_audio_chat"
18
  ],
19
  "_format_kwargs": {},
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "6342d438049fbc7e",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
 
14
  "wavllm_fairseq",
15
  "Qwen2-Audio-7B-Instruct",
16
  "whisper_large_v3_with_llama_3_8b_instruct",
17
+ "mowe_audio",
18
  "qwen_audio_chat"
19
  ],
20
  "_format_kwargs": {},
examples/ASR/Earnings21-Test/data-00000-of-00001.arrow CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:33dce5920f131621276ca79ef824fc52d5d5bb043d7672458fdbe037c7fb60fd
3
- size 310572568
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:026f3aba629a6ab6600e0cb70e057725aa5719dad0ed5710cac783974d3c7cb5
3
+ size 316326352
examples/ASR/Earnings21-Test/dataset_info.json CHANGED
@@ -110,6 +110,24 @@
110
  "_type": "Value"
111
  }
112
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
113
  "qwen_audio_chat": {
114
  "answer": {
115
  "dtype": "string",
 
110
  "_type": "Value"
111
  }
112
  },
113
+ "mowe_audio": {
114
+ "answer": {
115
+ "dtype": "string",
116
+ "_type": "Value"
117
+ },
118
+ "model_prediction": {
119
+ "dtype": "string",
120
+ "_type": "Value"
121
+ },
122
+ "task_type": {
123
+ "dtype": "string",
124
+ "_type": "Value"
125
+ },
126
+ "text": {
127
+ "dtype": "string",
128
+ "_type": "Value"
129
+ }
130
+ },
131
  "qwen_audio_chat": {
132
  "answer": {
133
  "dtype": "string",
examples/ASR/Earnings21-Test/sample_0.wav CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d8e88eccc61c5e8a5e62867c52a8ccfc4e108a5f48459f7df0eb1e9ae7f16d4f
3
- size 139072236
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ad501843c9718b06a1bac99177c5e85cb987c9b29a46f0da0970bced3e7df7a2
3
+ size 88692524
examples/ASR/Earnings21-Test/sample_1.wav CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:dcd3ce7847b7fb0f600bb3faeca3cd2a8f4992fd1de17bd831ef6ccfb2623f33
3
- size 55065644
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bd6ba77731011a6dc02e5854a600a2036713be4c2d71abf63fd6a89b86083c4f
3
+ size 178791280
examples/ASR/Earnings21-Test/sample_2.wav CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:607725d2a33f206a0f00668f1907f4d997fc7dbe7d1e1f9c82045496412bd8bd
3
- size 115039268
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:157daf8de352731745a22fbb4680f6d1b493ddfa31b1f80db4039f474a7d8ac0
3
+ size 47255852
examples/ASR/Earnings21-Test/state.json CHANGED
@@ -4,7 +4,7 @@
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
- "_fingerprint": "2e3dea299b387757",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
@@ -14,6 +14,7 @@
14
  "wavllm_fairseq",
15
  "Qwen2-Audio-7B-Instruct",
16
  "whisper_large_v3_with_llama_3_8b_instruct",
 
17
  "qwen_audio_chat"
18
  ],
19
  "_format_kwargs": {},
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "d0ad1703cbc51418",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
 
14
  "wavllm_fairseq",
15
  "Qwen2-Audio-7B-Instruct",
16
  "whisper_large_v3_with_llama_3_8b_instruct",
17
+ "mowe_audio",
18
  "qwen_audio_chat"
19
  ],
20
  "_format_kwargs": {},
examples/ASR/Earnings22-Test/data-00000-of-00001.arrow CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:51a45b6f2b6f1eaefc76f48f4778b5d5d46793e0a0fc846b1c2dd4b639dd173f
3
- size 299513312
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:80d673ef090f9c98675b12556bac6f7f17d1594e8c8c4a0ec81d069fb9921473
3
+ size 369181728
examples/ASR/Earnings22-Test/dataset_info.json CHANGED
@@ -110,6 +110,24 @@
110
  "_type": "Value"
111
  }
112
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
113
  "qwen_audio_chat": {
114
  "answer": {
115
  "dtype": "string",
 
110
  "_type": "Value"
111
  }
112
  },
113
+ "mowe_audio": {
114
+ "answer": {
115
+ "dtype": "string",
116
+ "_type": "Value"
117
+ },
118
+ "model_prediction": {
119
+ "dtype": "string",
120
+ "_type": "Value"
121
+ },
122
+ "task_type": {
123
+ "dtype": "string",
124
+ "_type": "Value"
125
+ },
126
+ "text": {
127
+ "dtype": "string",
128
+ "_type": "Value"
129
+ }
130
+ },
131
  "qwen_audio_chat": {
132
  "answer": {
133
  "dtype": "string",
examples/ASR/Earnings22-Test/sample_0.wav CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:df99676242890437003f9de1ebe2df3ea1e55365a76eab1b08bbe5ef6c235bf0
3
- size 39828524
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a9e7958f1ba96ac88655eab070583dadd0e380891c056b373a6dc6e7e631de75
3
+ size 129058604
examples/ASR/Earnings22-Test/sample_1.wav CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f46cd00f65ad5466bbe58fcbaa45dbbdc62921f876890b78da515b3574178691
3
- size 140347052
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7053d9e3d049cb299afcd1951678daf6706ada86c5edecdd208f0a5796b0c515
3
+ size 121073708
examples/ASR/Earnings22-Test/sample_2.wav CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d7144b858f3de344f601031d093c3403199dc04cad49287cf71dd6ea04ca4dd6
3
- size 118169900
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd5ca45b90a4997950deda21f13c95dbaf67cdcb23164df6fd7086c319e0d61e
3
+ size 117282092
examples/ASR/Earnings22-Test/state.json CHANGED
@@ -4,7 +4,7 @@
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
- "_fingerprint": "c2ddf91e8ccb230c",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
@@ -14,6 +14,7 @@
14
  "wavllm_fairseq",
15
  "Qwen2-Audio-7B-Instruct",
16
  "whisper_large_v3_with_llama_3_8b_instruct",
 
17
  "qwen_audio_chat"
18
  ],
19
  "_format_kwargs": {},
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "fb047ff90ed3a443",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
 
14
  "wavllm_fairseq",
15
  "Qwen2-Audio-7B-Instruct",
16
  "whisper_large_v3_with_llama_3_8b_instruct",
17
+ "mowe_audio",
18
  "qwen_audio_chat"
19
  ],
20
  "_format_kwargs": {},
examples/ASR/GigaSpeech-Test/data-00000-of-00001.arrow CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:573e5fd3cdcb0c5b27f9958ba68113a455fa426dec336f7635677c3130a265fe
3
- size 693672
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd97464fe4407bab4efb6550648ebd6b9d26e66198d5667411e67d7144eb0664
3
+ size 545808
examples/ASR/GigaSpeech-Test/dataset_info.json CHANGED
@@ -138,6 +138,24 @@
138
  "_type": "Value"
139
  }
140
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
141
  "qwen_audio_chat": {
142
  "answer": {
143
  "dtype": "string",
 
138
  "_type": "Value"
139
  }
140
  },
141
+ "mowe_audio": {
142
+ "answer": {
143
+ "dtype": "string",
144
+ "_type": "Value"
145
+ },
146
+ "model_prediction": {
147
+ "dtype": "string",
148
+ "_type": "Value"
149
+ },
150
+ "task_type": {
151
+ "dtype": "string",
152
+ "_type": "Value"
153
+ },
154
+ "text": {
155
+ "dtype": "string",
156
+ "_type": "Value"
157
+ }
158
+ },
159
  "qwen_audio_chat": {
160
  "answer": {
161
  "dtype": "string",
examples/ASR/GigaSpeech-Test/sample_0.wav CHANGED
Binary files a/examples/ASR/GigaSpeech-Test/sample_0.wav and b/examples/ASR/GigaSpeech-Test/sample_0.wav differ
 
examples/ASR/GigaSpeech-Test/sample_1.wav CHANGED
Binary files a/examples/ASR/GigaSpeech-Test/sample_1.wav and b/examples/ASR/GigaSpeech-Test/sample_1.wav differ
 
examples/ASR/GigaSpeech-Test/sample_2.wav CHANGED
Binary files a/examples/ASR/GigaSpeech-Test/sample_2.wav and b/examples/ASR/GigaSpeech-Test/sample_2.wav differ
 
examples/ASR/GigaSpeech-Test/state.json CHANGED
@@ -4,7 +4,7 @@
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
- "_fingerprint": "0032f92a85e94025",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
@@ -14,6 +14,7 @@
14
  "wavllm_fairseq",
15
  "Qwen2-Audio-7B-Instruct",
16
  "whisper_large_v3_with_llama_3_8b_instruct",
 
17
  "qwen_audio_chat"
18
  ],
19
  "_format_kwargs": {},
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "84a02614da440215",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
 
14
  "wavllm_fairseq",
15
  "Qwen2-Audio-7B-Instruct",
16
  "whisper_large_v3_with_llama_3_8b_instruct",
17
+ "mowe_audio",
18
  "qwen_audio_chat"
19
  ],
20
  "_format_kwargs": {},
examples/ASR/IMDA-Part1-ASR-Test/data-00000-of-00001.arrow CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:136cd0551a26eb0407715350491a49cd3c80cf8440c0c5d6b5079bdf1d81d260
3
- size 382752
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5f9d1180986a9e64590852464dcfb972170518d45f4fc0189b7fc3621b8e509e
3
+ size 485936
examples/ASR/IMDA-Part1-ASR-Test/sample_0.wav CHANGED
Binary files a/examples/ASR/IMDA-Part1-ASR-Test/sample_0.wav and b/examples/ASR/IMDA-Part1-ASR-Test/sample_0.wav differ
 
examples/ASR/IMDA-Part1-ASR-Test/sample_1.wav CHANGED
Binary files a/examples/ASR/IMDA-Part1-ASR-Test/sample_1.wav and b/examples/ASR/IMDA-Part1-ASR-Test/sample_1.wav differ
 
examples/ASR/IMDA-Part1-ASR-Test/sample_2.wav CHANGED
Binary files a/examples/ASR/IMDA-Part1-ASR-Test/sample_2.wav and b/examples/ASR/IMDA-Part1-ASR-Test/sample_2.wav differ
 
examples/ASR/IMDA-Part1-ASR-Test/state.json CHANGED
@@ -4,7 +4,7 @@
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
- "_fingerprint": "9b9c1437475afa9d",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "1514e693988caee7",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
examples/ASR/IMDA-Part2-ASR-Test/data-00000-of-00001.arrow CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ad482831507358784822a617955723251f237ff24e9dfba98265ebd1bebfacf5
3
- size 347952
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d8adfae25f905c40a6b1ccc95fa8f55826a0e6e8840f7b04e655aab873635941
3
+ size 328864
examples/ASR/IMDA-Part2-ASR-Test/sample_0.wav CHANGED
Binary files a/examples/ASR/IMDA-Part2-ASR-Test/sample_0.wav and b/examples/ASR/IMDA-Part2-ASR-Test/sample_0.wav differ
 
examples/ASR/IMDA-Part2-ASR-Test/sample_1.wav CHANGED
Binary files a/examples/ASR/IMDA-Part2-ASR-Test/sample_1.wav and b/examples/ASR/IMDA-Part2-ASR-Test/sample_1.wav differ
 
examples/ASR/IMDA-Part2-ASR-Test/sample_2.wav CHANGED
Binary files a/examples/ASR/IMDA-Part2-ASR-Test/sample_2.wav and b/examples/ASR/IMDA-Part2-ASR-Test/sample_2.wav differ
 
examples/ASR/IMDA-Part2-ASR-Test/state.json CHANGED
@@ -4,7 +4,7 @@
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
- "_fingerprint": "1b048a2e54d0c002",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "8a8e117080f24a8b",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
examples/ASR/LibriSpeech-Test-Clean/data-00000-of-00001.arrow CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:28a1a86f5cd34ee42918c78eef50f95a53953fd352fb0ecc758543025634d2b7
3
- size 555160
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:af079564f35d59eb2cbfd4a1dba6dec7692bd78025e45de8d3915575a60218ca
3
+ size 747384
examples/ASR/LibriSpeech-Test-Clean/dataset_info.json CHANGED
@@ -122,6 +122,24 @@
122
  "_type": "Value"
123
  }
124
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
125
  "qwen_audio_chat": {
126
  "answer": {
127
  "dtype": "string",
 
122
  "_type": "Value"
123
  }
124
  },
125
+ "mowe_audio": {
126
+ "answer": {
127
+ "dtype": "string",
128
+ "_type": "Value"
129
+ },
130
+ "model_prediction": {
131
+ "dtype": "string",
132
+ "_type": "Value"
133
+ },
134
+ "task_type": {
135
+ "dtype": "string",
136
+ "_type": "Value"
137
+ },
138
+ "text": {
139
+ "dtype": "string",
140
+ "_type": "Value"
141
+ }
142
+ },
143
  "qwen_audio_chat": {
144
  "answer": {
145
  "dtype": "string",
examples/ASR/LibriSpeech-Test-Clean/sample_0.wav CHANGED
Binary files a/examples/ASR/LibriSpeech-Test-Clean/sample_0.wav and b/examples/ASR/LibriSpeech-Test-Clean/sample_0.wav differ