Upload folder using huggingface_hub
Browse filesThis view is limited to 50 files because it contains too many changes.
See raw diff
- .gitattributes +4 -0
- examples/AC/AudioCaps-Test/sample_0.wav +0 -0
- examples/AC/AudioCaps-Test/sample_1.wav +0 -0
- examples/AC/AudioCaps-Test/sample_2.wav +0 -0
- examples/AC/AudioCaps-Test/state.json +1 -1
- examples/AC/WavCaps-Test/sample_0.wav +0 -0
- examples/AC/WavCaps-Test/sample_1.wav +0 -0
- examples/AC/WavCaps-Test/sample_2.wav +0 -0
- examples/AC/WavCaps-Test/state.json +1 -1
- examples/AQA/AudioCaps-QA-Test/sample_0.wav +0 -0
- examples/AQA/AudioCaps-QA-Test/sample_1.wav +0 -0
- examples/AQA/AudioCaps-QA-Test/sample_2.wav +0 -0
- examples/AQA/AudioCaps-QA-Test/state.json +1 -1
- examples/AQA/Clotho-AQA-Test/dataset_info.json +109 -1
- examples/AQA/Clotho-AQA-Test/sample_0.wav +0 -0
- examples/AQA/Clotho-AQA-Test/sample_1.wav +0 -0
- examples/AQA/Clotho-AQA-Test/sample_2.wav +0 -0
- examples/AQA/Clotho-AQA-Test/state.json +8 -2
- examples/AQA/WavCaps-QA-Test/sample_0.wav +0 -0
- examples/AQA/WavCaps-QA-Test/sample_1.wav +0 -0
- examples/AQA/WavCaps-QA-Test/sample_2.wav +0 -0
- examples/AQA/WavCaps-QA-Test/state.json +1 -1
- examples/AR/VoxCeleb-Accent-Test/sample_0.wav +0 -0
- examples/AR/VoxCeleb-Accent-Test/sample_1.wav +0 -0
- examples/AR/VoxCeleb-Accent-Test/sample_2.wav +0 -0
- examples/AR/VoxCeleb-Accent-Test/state.json +1 -1
- examples/ASR/Common-Voice-15-En-Test/sample_0.wav +0 -0
- examples/ASR/Common-Voice-15-En-Test/sample_1.wav +0 -0
- examples/ASR/Common-Voice-15-En-Test/sample_2.wav +0 -0
- examples/ASR/Common-Voice-15-En-Test/state.json +1 -1
- examples/ASR/Earnings21-Test/state.json +1 -1
- examples/ASR/Earnings22-Test/state.json +1 -1
- examples/ASR/GigaSpeech-Test/sample_0.wav +0 -0
- examples/ASR/GigaSpeech-Test/sample_1.wav +0 -0
- examples/ASR/GigaSpeech-Test/sample_2.wav +0 -0
- examples/ASR/GigaSpeech-Test/state.json +1 -1
- examples/ASR/IMDA-Part1-ASR-Test/dataset_info.json +0 -72
- examples/ASR/IMDA-Part1-ASR-Test/sample_0.wav +0 -0
- examples/ASR/IMDA-Part1-ASR-Test/sample_1.wav +0 -0
- examples/ASR/IMDA-Part1-ASR-Test/sample_2.wav +0 -0
- examples/ASR/IMDA-Part1-ASR-Test/state.json +2 -6
- examples/ASR/IMDA-Part2-ASR-Test/dataset_info.json +0 -72
- examples/ASR/IMDA-Part2-ASR-Test/sample_0.wav +0 -0
- examples/ASR/IMDA-Part2-ASR-Test/sample_1.wav +0 -0
- examples/ASR/IMDA-Part2-ASR-Test/sample_2.wav +0 -0
- examples/ASR/IMDA-Part2-ASR-Test/state.json +2 -6
- examples/ASR/LibriSpeech-Test-Clean/sample_0.wav +0 -0
- examples/ASR/LibriSpeech-Test-Clean/sample_1.wav +0 -0
- examples/ASR/LibriSpeech-Test-Clean/sample_2.wav +0 -0
- examples/ASR/LibriSpeech-Test-Clean/state.json +1 -1
.gitattributes
CHANGED
|
@@ -56,3 +56,7 @@ examples/ASR/Tedlium3-Long-form-Test/sample_0.wav filter=lfs diff=lfs merge=lfs
|
|
| 56 |
examples/ASR/Tedlium3-Long-form-Test/sample_1.wav filter=lfs diff=lfs merge=lfs -text
|
| 57 |
examples/ASR/Tedlium3-Long-form-Test/sample_2.wav filter=lfs diff=lfs merge=lfs -text
|
| 58 |
examples/SQA/Public-SG-Speech-QA-Test/sample_2.wav filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 56 |
examples/ASR/Tedlium3-Long-form-Test/sample_1.wav filter=lfs diff=lfs merge=lfs -text
|
| 57 |
examples/ASR/Tedlium3-Long-form-Test/sample_2.wav filter=lfs diff=lfs merge=lfs -text
|
| 58 |
examples/SQA/Public-SG-Speech-QA-Test/sample_2.wav filter=lfs diff=lfs merge=lfs -text
|
| 59 |
+
examples/SQA/CN-College-Listen-MCQ-Test/sample_0.wav filter=lfs diff=lfs merge=lfs -text
|
| 60 |
+
examples/SQA/Spoken-Squad-Test/sample_0.wav filter=lfs diff=lfs merge=lfs -text
|
| 61 |
+
examples/SQA/Spoken-Squad-Test/sample_1.wav filter=lfs diff=lfs merge=lfs -text
|
| 62 |
+
examples/SQA/Spoken-Squad-Test/sample_2.wav filter=lfs diff=lfs merge=lfs -text
|
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": "
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
|
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
+
"_fingerprint": "e736bf1821a473f3",
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
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": "
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
|
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
+
"_fingerprint": "742ab313af054565",
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
examples/AQA/AudioCaps-QA-Test/sample_0.wav
CHANGED
|
Binary files a/examples/AQA/AudioCaps-QA-Test/sample_0.wav and b/examples/AQA/AudioCaps-QA-Test/sample_0.wav differ
|
|
|
examples/AQA/AudioCaps-QA-Test/sample_1.wav
CHANGED
|
Binary files a/examples/AQA/AudioCaps-QA-Test/sample_1.wav and b/examples/AQA/AudioCaps-QA-Test/sample_1.wav differ
|
|
|
examples/AQA/AudioCaps-QA-Test/sample_2.wav
CHANGED
|
Binary files a/examples/AQA/AudioCaps-QA-Test/sample_2.wav and b/examples/AQA/AudioCaps-QA-Test/sample_2.wav differ
|
|
|
examples/AQA/AudioCaps-QA-Test/state.json
CHANGED
|
@@ -4,7 +4,7 @@
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
-
"_fingerprint": "
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
|
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
+
"_fingerprint": "52bc1dfcaf2a0f4b",
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
examples/AQA/Clotho-AQA-Test/dataset_info.json
CHANGED
|
@@ -32,7 +32,115 @@
|
|
| 32 |
"_type": "Value"
|
| 33 |
}
|
| 34 |
},
|
| 35 |
-
"other_attributes": {}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
},
|
| 37 |
"homepage": "",
|
| 38 |
"license": ""
|
|
|
|
| 32 |
"_type": "Value"
|
| 33 |
}
|
| 34 |
},
|
| 35 |
+
"other_attributes": {},
|
| 36 |
+
"salmonn_7b": {
|
| 37 |
+
"answer": {
|
| 38 |
+
"dtype": "string",
|
| 39 |
+
"_type": "Value"
|
| 40 |
+
},
|
| 41 |
+
"model_prediction": {
|
| 42 |
+
"dtype": "string",
|
| 43 |
+
"_type": "Value"
|
| 44 |
+
},
|
| 45 |
+
"task_type": {
|
| 46 |
+
"dtype": "string",
|
| 47 |
+
"_type": "Value"
|
| 48 |
+
},
|
| 49 |
+
"text": {
|
| 50 |
+
"dtype": "string",
|
| 51 |
+
"_type": "Value"
|
| 52 |
+
}
|
| 53 |
+
},
|
| 54 |
+
"wavllm_fairseq": {
|
| 55 |
+
"answer": {
|
| 56 |
+
"dtype": "string",
|
| 57 |
+
"_type": "Value"
|
| 58 |
+
},
|
| 59 |
+
"model_prediction": {
|
| 60 |
+
"dtype": "string",
|
| 61 |
+
"_type": "Value"
|
| 62 |
+
},
|
| 63 |
+
"task_type": {
|
| 64 |
+
"dtype": "string",
|
| 65 |
+
"_type": "Value"
|
| 66 |
+
},
|
| 67 |
+
"text": {
|
| 68 |
+
"dtype": "string",
|
| 69 |
+
"_type": "Value"
|
| 70 |
+
}
|
| 71 |
+
},
|
| 72 |
+
"Qwen2-Audio-7B-Instruct": {
|
| 73 |
+
"answer": {
|
| 74 |
+
"dtype": "string",
|
| 75 |
+
"_type": "Value"
|
| 76 |
+
},
|
| 77 |
+
"model_prediction": {
|
| 78 |
+
"dtype": "string",
|
| 79 |
+
"_type": "Value"
|
| 80 |
+
},
|
| 81 |
+
"task_type": {
|
| 82 |
+
"dtype": "string",
|
| 83 |
+
"_type": "Value"
|
| 84 |
+
},
|
| 85 |
+
"text": {
|
| 86 |
+
"dtype": "string",
|
| 87 |
+
"_type": "Value"
|
| 88 |
+
}
|
| 89 |
+
},
|
| 90 |
+
"whisper_large_v3_with_llama_3_8b_instruct": {
|
| 91 |
+
"answer": {
|
| 92 |
+
"dtype": "string",
|
| 93 |
+
"_type": "Value"
|
| 94 |
+
},
|
| 95 |
+
"model_prediction": {
|
| 96 |
+
"dtype": "string",
|
| 97 |
+
"_type": "Value"
|
| 98 |
+
},
|
| 99 |
+
"task_type": {
|
| 100 |
+
"dtype": "string",
|
| 101 |
+
"_type": "Value"
|
| 102 |
+
},
|
| 103 |
+
"text": {
|
| 104 |
+
"dtype": "string",
|
| 105 |
+
"_type": "Value"
|
| 106 |
+
}
|
| 107 |
+
},
|
| 108 |
+
"mowe_audio": {
|
| 109 |
+
"answer": {
|
| 110 |
+
"dtype": "string",
|
| 111 |
+
"_type": "Value"
|
| 112 |
+
},
|
| 113 |
+
"model_prediction": {
|
| 114 |
+
"dtype": "string",
|
| 115 |
+
"_type": "Value"
|
| 116 |
+
},
|
| 117 |
+
"task_type": {
|
| 118 |
+
"dtype": "string",
|
| 119 |
+
"_type": "Value"
|
| 120 |
+
},
|
| 121 |
+
"text": {
|
| 122 |
+
"dtype": "string",
|
| 123 |
+
"_type": "Value"
|
| 124 |
+
}
|
| 125 |
+
},
|
| 126 |
+
"qwen_audio_chat": {
|
| 127 |
+
"answer": {
|
| 128 |
+
"dtype": "string",
|
| 129 |
+
"_type": "Value"
|
| 130 |
+
},
|
| 131 |
+
"model_prediction": {
|
| 132 |
+
"dtype": "string",
|
| 133 |
+
"_type": "Value"
|
| 134 |
+
},
|
| 135 |
+
"task_type": {
|
| 136 |
+
"dtype": "string",
|
| 137 |
+
"_type": "Value"
|
| 138 |
+
},
|
| 139 |
+
"text": {
|
| 140 |
+
"dtype": "string",
|
| 141 |
+
"_type": "Value"
|
| 142 |
+
}
|
| 143 |
+
}
|
| 144 |
},
|
| 145 |
"homepage": "",
|
| 146 |
"license": ""
|
examples/AQA/Clotho-AQA-Test/sample_0.wav
CHANGED
|
Binary files a/examples/AQA/Clotho-AQA-Test/sample_0.wav and b/examples/AQA/Clotho-AQA-Test/sample_0.wav differ
|
|
|
examples/AQA/Clotho-AQA-Test/sample_1.wav
CHANGED
|
Binary files a/examples/AQA/Clotho-AQA-Test/sample_1.wav and b/examples/AQA/Clotho-AQA-Test/sample_1.wav differ
|
|
|
examples/AQA/Clotho-AQA-Test/sample_2.wav
CHANGED
|
Binary files a/examples/AQA/Clotho-AQA-Test/sample_2.wav and b/examples/AQA/Clotho-AQA-Test/sample_2.wav differ
|
|
|
examples/AQA/Clotho-AQA-Test/state.json
CHANGED
|
@@ -4,12 +4,18 @@
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
-
"_fingerprint": "
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
| 11 |
"answer",
|
| 12 |
-
"other_attributes"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
],
|
| 14 |
"_format_kwargs": {},
|
| 15 |
"_format_type": null,
|
|
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
+
"_fingerprint": "e2e76326f448d7c4",
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
| 11 |
"answer",
|
| 12 |
+
"other_attributes",
|
| 13 |
+
"salmonn_7b",
|
| 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": {},
|
| 21 |
"_format_type": null,
|
examples/AQA/WavCaps-QA-Test/sample_0.wav
CHANGED
|
Binary files a/examples/AQA/WavCaps-QA-Test/sample_0.wav and b/examples/AQA/WavCaps-QA-Test/sample_0.wav differ
|
|
|
examples/AQA/WavCaps-QA-Test/sample_1.wav
CHANGED
|
Binary files a/examples/AQA/WavCaps-QA-Test/sample_1.wav and b/examples/AQA/WavCaps-QA-Test/sample_1.wav differ
|
|
|
examples/AQA/WavCaps-QA-Test/sample_2.wav
CHANGED
|
Binary files a/examples/AQA/WavCaps-QA-Test/sample_2.wav and b/examples/AQA/WavCaps-QA-Test/sample_2.wav differ
|
|
|
examples/AQA/WavCaps-QA-Test/state.json
CHANGED
|
@@ -4,7 +4,7 @@
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
-
"_fingerprint": "
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
|
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
+
"_fingerprint": "40995a6cc1fe3dc7",
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
examples/AR/VoxCeleb-Accent-Test/sample_0.wav
CHANGED
|
Binary files a/examples/AR/VoxCeleb-Accent-Test/sample_0.wav and b/examples/AR/VoxCeleb-Accent-Test/sample_0.wav differ
|
|
|
examples/AR/VoxCeleb-Accent-Test/sample_1.wav
CHANGED
|
Binary files a/examples/AR/VoxCeleb-Accent-Test/sample_1.wav and b/examples/AR/VoxCeleb-Accent-Test/sample_1.wav differ
|
|
|
examples/AR/VoxCeleb-Accent-Test/sample_2.wav
CHANGED
|
Binary files a/examples/AR/VoxCeleb-Accent-Test/sample_2.wav and b/examples/AR/VoxCeleb-Accent-Test/sample_2.wav differ
|
|
|
examples/AR/VoxCeleb-Accent-Test/state.json
CHANGED
|
@@ -4,7 +4,7 @@
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
-
"_fingerprint": "
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
|
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
+
"_fingerprint": "fa91a59f90c22c3c",
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
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": "
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
|
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
+
"_fingerprint": "468db91ad949e4d4",
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
examples/ASR/Earnings21-Test/state.json
CHANGED
|
@@ -4,7 +4,7 @@
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
-
"_fingerprint": "
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
|
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
+
"_fingerprint": "8cc0ad99446f1aba",
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
examples/ASR/Earnings22-Test/state.json
CHANGED
|
@@ -4,7 +4,7 @@
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
-
"_fingerprint": "
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
|
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
+
"_fingerprint": "331c061bce6e651c",
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
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": "
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
|
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
+
"_fingerprint": "68d371cc267ff1d2",
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
examples/ASR/IMDA-Part1-ASR-Test/dataset_info.json
CHANGED
|
@@ -85,78 +85,6 @@
|
|
| 85 |
"_type": "Value"
|
| 86 |
}
|
| 87 |
}
|
| 88 |
-
},
|
| 89 |
-
"salmonn_7b": {
|
| 90 |
-
"answer": {
|
| 91 |
-
"dtype": "string",
|
| 92 |
-
"_type": "Value"
|
| 93 |
-
},
|
| 94 |
-
"model_prediction": {
|
| 95 |
-
"dtype": "string",
|
| 96 |
-
"_type": "Value"
|
| 97 |
-
},
|
| 98 |
-
"task_type": {
|
| 99 |
-
"dtype": "string",
|
| 100 |
-
"_type": "Value"
|
| 101 |
-
},
|
| 102 |
-
"text": {
|
| 103 |
-
"dtype": "string",
|
| 104 |
-
"_type": "Value"
|
| 105 |
-
}
|
| 106 |
-
},
|
| 107 |
-
"wavllm_fairseq": {
|
| 108 |
-
"answer": {
|
| 109 |
-
"dtype": "string",
|
| 110 |
-
"_type": "Value"
|
| 111 |
-
},
|
| 112 |
-
"model_prediction": {
|
| 113 |
-
"dtype": "string",
|
| 114 |
-
"_type": "Value"
|
| 115 |
-
},
|
| 116 |
-
"task_type": {
|
| 117 |
-
"dtype": "string",
|
| 118 |
-
"_type": "Value"
|
| 119 |
-
},
|
| 120 |
-
"text": {
|
| 121 |
-
"dtype": "string",
|
| 122 |
-
"_type": "Value"
|
| 123 |
-
}
|
| 124 |
-
},
|
| 125 |
-
"whisper_large_v3_with_llama_3_8b_instruct": {
|
| 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",
|
| 146 |
-
"_type": "Value"
|
| 147 |
-
},
|
| 148 |
-
"model_prediction": {
|
| 149 |
-
"dtype": "string",
|
| 150 |
-
"_type": "Value"
|
| 151 |
-
},
|
| 152 |
-
"task_type": {
|
| 153 |
-
"dtype": "string",
|
| 154 |
-
"_type": "Value"
|
| 155 |
-
},
|
| 156 |
-
"text": {
|
| 157 |
-
"dtype": "string",
|
| 158 |
-
"_type": "Value"
|
| 159 |
-
}
|
| 160 |
}
|
| 161 |
},
|
| 162 |
"homepage": "",
|
|
|
|
| 85 |
"_type": "Value"
|
| 86 |
}
|
| 87 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 88 |
}
|
| 89 |
},
|
| 90 |
"homepage": "",
|
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
|
@@ -4,16 +4,12 @@
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
-
"_fingerprint": "
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
| 11 |
"answer",
|
| 12 |
-
"other_attributes"
|
| 13 |
-
"salmonn_7b",
|
| 14 |
-
"wavllm_fairseq",
|
| 15 |
-
"whisper_large_v3_with_llama_3_8b_instruct",
|
| 16 |
-
"qwen_audio_chat"
|
| 17 |
],
|
| 18 |
"_format_kwargs": {},
|
| 19 |
"_format_type": null,
|
|
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
+
"_fingerprint": "9c67b4f2f347692a",
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
| 11 |
"answer",
|
| 12 |
+
"other_attributes"
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
],
|
| 14 |
"_format_kwargs": {},
|
| 15 |
"_format_type": null,
|
examples/ASR/IMDA-Part2-ASR-Test/dataset_info.json
CHANGED
|
@@ -85,78 +85,6 @@
|
|
| 85 |
"_type": "Value"
|
| 86 |
}
|
| 87 |
}
|
| 88 |
-
},
|
| 89 |
-
"salmonn_7b": {
|
| 90 |
-
"answer": {
|
| 91 |
-
"dtype": "string",
|
| 92 |
-
"_type": "Value"
|
| 93 |
-
},
|
| 94 |
-
"model_prediction": {
|
| 95 |
-
"dtype": "string",
|
| 96 |
-
"_type": "Value"
|
| 97 |
-
},
|
| 98 |
-
"task_type": {
|
| 99 |
-
"dtype": "string",
|
| 100 |
-
"_type": "Value"
|
| 101 |
-
},
|
| 102 |
-
"text": {
|
| 103 |
-
"dtype": "string",
|
| 104 |
-
"_type": "Value"
|
| 105 |
-
}
|
| 106 |
-
},
|
| 107 |
-
"wavllm_fairseq": {
|
| 108 |
-
"answer": {
|
| 109 |
-
"dtype": "string",
|
| 110 |
-
"_type": "Value"
|
| 111 |
-
},
|
| 112 |
-
"model_prediction": {
|
| 113 |
-
"dtype": "string",
|
| 114 |
-
"_type": "Value"
|
| 115 |
-
},
|
| 116 |
-
"task_type": {
|
| 117 |
-
"dtype": "string",
|
| 118 |
-
"_type": "Value"
|
| 119 |
-
},
|
| 120 |
-
"text": {
|
| 121 |
-
"dtype": "string",
|
| 122 |
-
"_type": "Value"
|
| 123 |
-
}
|
| 124 |
-
},
|
| 125 |
-
"whisper_large_v3_with_llama_3_8b_instruct": {
|
| 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",
|
| 146 |
-
"_type": "Value"
|
| 147 |
-
},
|
| 148 |
-
"model_prediction": {
|
| 149 |
-
"dtype": "string",
|
| 150 |
-
"_type": "Value"
|
| 151 |
-
},
|
| 152 |
-
"task_type": {
|
| 153 |
-
"dtype": "string",
|
| 154 |
-
"_type": "Value"
|
| 155 |
-
},
|
| 156 |
-
"text": {
|
| 157 |
-
"dtype": "string",
|
| 158 |
-
"_type": "Value"
|
| 159 |
-
}
|
| 160 |
}
|
| 161 |
},
|
| 162 |
"homepage": "",
|
|
|
|
| 85 |
"_type": "Value"
|
| 86 |
}
|
| 87 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 88 |
}
|
| 89 |
},
|
| 90 |
"homepage": "",
|
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
|
@@ -4,16 +4,12 @@
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
-
"_fingerprint": "
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
| 11 |
"answer",
|
| 12 |
-
"other_attributes"
|
| 13 |
-
"salmonn_7b",
|
| 14 |
-
"wavllm_fairseq",
|
| 15 |
-
"whisper_large_v3_with_llama_3_8b_instruct",
|
| 16 |
-
"qwen_audio_chat"
|
| 17 |
],
|
| 18 |
"_format_kwargs": {},
|
| 19 |
"_format_type": null,
|
|
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
+
"_fingerprint": "9f3d440792a605d2",
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
| 11 |
"answer",
|
| 12 |
+
"other_attributes"
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
],
|
| 14 |
"_format_kwargs": {},
|
| 15 |
"_format_type": null,
|
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
|
|
|
examples/ASR/LibriSpeech-Test-Clean/sample_1.wav
CHANGED
|
Binary files a/examples/ASR/LibriSpeech-Test-Clean/sample_1.wav and b/examples/ASR/LibriSpeech-Test-Clean/sample_1.wav differ
|
|
|
examples/ASR/LibriSpeech-Test-Clean/sample_2.wav
CHANGED
|
Binary files a/examples/ASR/LibriSpeech-Test-Clean/sample_2.wav and b/examples/ASR/LibriSpeech-Test-Clean/sample_2.wav differ
|
|
|
examples/ASR/LibriSpeech-Test-Clean/state.json
CHANGED
|
@@ -4,7 +4,7 @@
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
-
"_fingerprint": "
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|
|
|
|
| 4 |
"filename": "data-00000-of-00001.arrow"
|
| 5 |
}
|
| 6 |
],
|
| 7 |
+
"_fingerprint": "d9f5d173c305ae96",
|
| 8 |
"_format_columns": [
|
| 9 |
"context",
|
| 10 |
"instruction",
|