Hugging Face's logo

Spaces:
MERaLiON
/
AudioBench-Leaderboard
Running

App Files Community
1
zhuohan-7 commited on
Commit
12b814c
·
1 Parent(s): a7c39e1

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. examples/AC/AudioCaps-Test/dataset_info.json +36 -0
  2. examples/AC/AudioCaps-Test/sample_0.wav +0 -0
  3. examples/AC/AudioCaps-Test/sample_1.wav +0 -0
  4. examples/AC/AudioCaps-Test/sample_2.wav +0 -0
  5. examples/AC/AudioCaps-Test/state.json +4 -2
  6. examples/AC/WavCaps-Test/dataset_info.json +36 -0
  7. examples/AC/WavCaps-Test/sample_0.wav +0 -0
  8. examples/AC/WavCaps-Test/sample_1.wav +0 -0
  9. examples/AC/WavCaps-Test/sample_2.wav +0 -0
  10. examples/AC/WavCaps-Test/state.json +4 -2
  11. examples/AQA/AudioCaps-QA-Test/dataset_info.json +36 -0
  12. examples/AQA/AudioCaps-QA-Test/sample_0.wav +0 -0
  13. examples/AQA/AudioCaps-QA-Test/sample_1.wav +0 -0
  14. examples/AQA/AudioCaps-QA-Test/sample_2.wav +0 -0
  15. examples/AQA/AudioCaps-QA-Test/state.json +4 -2
  16. examples/AQA/Clotho-AQA-Test/dataset_info.json +36 -0
  17. examples/AQA/Clotho-AQA-Test/sample_0.wav +0 -0
  18. examples/AQA/Clotho-AQA-Test/sample_1.wav +0 -0
  19. examples/AQA/Clotho-AQA-Test/sample_2.wav +0 -0
  20. examples/AQA/Clotho-AQA-Test/state.json +4 -2
  21. examples/AQA/WavCaps-QA-Test/dataset_info.json +36 -0
  22. examples/AQA/WavCaps-QA-Test/sample_0.wav +0 -0
  23. examples/AQA/WavCaps-QA-Test/sample_1.wav +0 -0
  24. examples/AQA/WavCaps-QA-Test/sample_2.wav +0 -0
  25. examples/AQA/WavCaps-QA-Test/state.json +4 -2
  26. examples/AR/VoxCeleb-Accent-Test/dataset_info.json +36 -0
  27. examples/AR/VoxCeleb-Accent-Test/sample_0.wav +0 -0
  28. examples/AR/VoxCeleb-Accent-Test/sample_1.wav +0 -0
  29. examples/AR/VoxCeleb-Accent-Test/sample_2.wav +0 -0
  30. examples/AR/VoxCeleb-Accent-Test/state.json +4 -2
  31. examples/ASR/Common-Voice-15-En-Test/dataset_info.json +37 -1
  32. examples/ASR/Common-Voice-15-En-Test/sample_0.wav +0 -0
  33. examples/ASR/Common-Voice-15-En-Test/sample_1.wav +0 -0
  34. examples/ASR/Common-Voice-15-En-Test/sample_2.wav +0 -0
  35. examples/ASR/Common-Voice-15-En-Test/state.json +4 -2
  36. examples/ASR/Earnings21-Test/dataset_info.json +36 -0
  37. examples/ASR/Earnings21-Test/state.json +4 -2
  38. examples/ASR/Earnings22-Test/dataset_info.json +54 -0
  39. examples/ASR/Earnings22-Test/state.json +5 -2
  40. examples/ASR/GigaSpeech-Test/dataset_info.json +36 -0
  41. examples/ASR/GigaSpeech-Test/sample_0.wav +0 -0
  42. examples/ASR/GigaSpeech-Test/sample_1.wav +0 -0
  43. examples/ASR/GigaSpeech-Test/sample_2.wav +0 -0
  44. examples/ASR/GigaSpeech-Test/state.json +4 -2
  45. examples/ASR/IMDA-Part1-ASR-Test/dataset_info.json +36 -0
  46. examples/ASR/IMDA-Part1-ASR-Test/sample_0.wav +0 -0
  47. examples/ASR/IMDA-Part1-ASR-Test/sample_1.wav +0 -0
  48. examples/ASR/IMDA-Part1-ASR-Test/sample_2.wav +0 -0
  49. examples/ASR/IMDA-Part1-ASR-Test/state.json +4 -2
  50. examples/ASR/IMDA-Part2-ASR-Test/dataset_info.json +36 -0
examples/AC/AudioCaps-Test/dataset_info.json CHANGED
@@ -171,6 +171,42 @@
171
  "dtype": "string",
172
  "_type": "Value"
173
  }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
174
  }
175
  },
176
  "homepage": "",
 
171
  "dtype": "string",
172
  "_type": "Value"
173
  }
174
+ },
175
+ "meralion_audiollm_v1_mse": {
176
+ "answer": {
177
+ "dtype": "string",
178
+ "_type": "Value"
179
+ },
180
+ "model_prediction": {
181
+ "dtype": "string",
182
+ "_type": "Value"
183
+ },
184
+ "task_type": {
185
+ "dtype": "string",
186
+ "_type": "Value"
187
+ },
188
+ "text": {
189
+ "dtype": "string",
190
+ "_type": "Value"
191
+ }
192
+ },
193
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora": {
194
+ "answer": {
195
+ "dtype": "string",
196
+ "_type": "Value"
197
+ },
198
+ "model_prediction": {
199
+ "dtype": "string",
200
+ "_type": "Value"
201
+ },
202
+ "task_type": {
203
+ "dtype": "string",
204
+ "_type": "Value"
205
+ },
206
+ "text": {
207
+ "dtype": "string",
208
+ "_type": "Value"
209
+ }
210
  }
211
  },
212
  "homepage": "",
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": "7dd956b95601f713",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
@@ -16,7 +16,9 @@
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
  "mowe_audio",
19
- "qwen_audio_chat"
 
 
20
  ],
21
  "_format_kwargs": {},
22
  "_format_type": null,
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "7cc5b61ce18c8217",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
 
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
  "mowe_audio",
19
+ "qwen_audio_chat",
20
+ "meralion_audiollm_v1_mse",
21
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora"
22
  ],
23
  "_format_kwargs": {},
24
  "_format_type": null,
examples/AC/WavCaps-Test/dataset_info.json CHANGED
@@ -167,6 +167,42 @@
167
  "dtype": "string",
168
  "_type": "Value"
169
  }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
170
  }
171
  },
172
  "homepage": "",
 
167
  "dtype": "string",
168
  "_type": "Value"
169
  }
170
+ },
171
+ "meralion_audiollm_v1_mse": {
172
+ "answer": {
173
+ "dtype": "string",
174
+ "_type": "Value"
175
+ },
176
+ "model_prediction": {
177
+ "dtype": "string",
178
+ "_type": "Value"
179
+ },
180
+ "task_type": {
181
+ "dtype": "string",
182
+ "_type": "Value"
183
+ },
184
+ "text": {
185
+ "dtype": "string",
186
+ "_type": "Value"
187
+ }
188
+ },
189
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora": {
190
+ "answer": {
191
+ "dtype": "string",
192
+ "_type": "Value"
193
+ },
194
+ "model_prediction": {
195
+ "dtype": "string",
196
+ "_type": "Value"
197
+ },
198
+ "task_type": {
199
+ "dtype": "string",
200
+ "_type": "Value"
201
+ },
202
+ "text": {
203
+ "dtype": "string",
204
+ "_type": "Value"
205
+ }
206
  }
207
  },
208
  "homepage": "",
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": "22a6dfe54867e49c",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
@@ -16,7 +16,9 @@
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
  "mowe_audio",
19
- "qwen_audio_chat"
 
 
20
  ],
21
  "_format_kwargs": {},
22
  "_format_type": null,
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "c7a1810f866a86af",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
 
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
  "mowe_audio",
19
+ "qwen_audio_chat",
20
+ "meralion_audiollm_v1_mse",
21
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora"
22
  ],
23
  "_format_kwargs": {},
24
  "_format_type": null,
examples/AQA/AudioCaps-QA-Test/dataset_info.json CHANGED
@@ -175,6 +175,42 @@
175
  "dtype": "string",
176
  "_type": "Value"
177
  }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
178
  }
179
  },
180
  "homepage": "",
 
175
  "dtype": "string",
176
  "_type": "Value"
177
  }
178
+ },
179
+ "meralion_audiollm_v1_mse": {
180
+ "answer": {
181
+ "dtype": "string",
182
+ "_type": "Value"
183
+ },
184
+ "model_prediction": {
185
+ "dtype": "string",
186
+ "_type": "Value"
187
+ },
188
+ "task_type": {
189
+ "dtype": "string",
190
+ "_type": "Value"
191
+ },
192
+ "text": {
193
+ "dtype": "string",
194
+ "_type": "Value"
195
+ }
196
+ },
197
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora": {
198
+ "answer": {
199
+ "dtype": "string",
200
+ "_type": "Value"
201
+ },
202
+ "model_prediction": {
203
+ "dtype": "string",
204
+ "_type": "Value"
205
+ },
206
+ "task_type": {
207
+ "dtype": "string",
208
+ "_type": "Value"
209
+ },
210
+ "text": {
211
+ "dtype": "string",
212
+ "_type": "Value"
213
+ }
214
  }
215
  },
216
  "homepage": "",
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": "60b01046f3ad5343",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
@@ -16,7 +16,9 @@
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
  "mowe_audio",
19
- "qwen_audio_chat"
 
 
20
  ],
21
  "_format_kwargs": {},
22
  "_format_type": null,
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "15709330a46e2556",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
 
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
  "mowe_audio",
19
+ "qwen_audio_chat",
20
+ "meralion_audiollm_v1_mse",
21
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora"
22
  ],
23
  "_format_kwargs": {},
24
  "_format_type": null,
examples/AQA/Clotho-AQA-Test/dataset_info.json CHANGED
@@ -158,6 +158,42 @@
158
  "dtype": "string",
159
  "_type": "Value"
160
  }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
161
  }
162
  },
163
  "homepage": "",
 
158
  "dtype": "string",
159
  "_type": "Value"
160
  }
161
+ },
162
+ "meralion_audiollm_v1_mse": {
163
+ "answer": {
164
+ "dtype": "string",
165
+ "_type": "Value"
166
+ },
167
+ "model_prediction": {
168
+ "dtype": "string",
169
+ "_type": "Value"
170
+ },
171
+ "task_type": {
172
+ "dtype": "string",
173
+ "_type": "Value"
174
+ },
175
+ "text": {
176
+ "dtype": "string",
177
+ "_type": "Value"
178
+ }
179
+ },
180
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora": {
181
+ "answer": {
182
+ "dtype": "string",
183
+ "_type": "Value"
184
+ },
185
+ "model_prediction": {
186
+ "dtype": "string",
187
+ "_type": "Value"
188
+ },
189
+ "task_type": {
190
+ "dtype": "string",
191
+ "_type": "Value"
192
+ },
193
+ "text": {
194
+ "dtype": "string",
195
+ "_type": "Value"
196
+ }
197
  }
198
  },
199
  "homepage": "",
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
@@ -4,7 +4,7 @@
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
- "_fingerprint": "9728812a68aca05b",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
@@ -16,7 +16,9 @@
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
  "mowe_audio",
19
- "qwen_audio_chat"
 
 
20
  ],
21
  "_format_kwargs": {},
22
  "_format_type": null,
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "185dd63ad411b41a",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
 
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
  "mowe_audio",
19
+ "qwen_audio_chat",
20
+ "meralion_audiollm_v1_mse",
21
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora"
22
  ],
23
  "_format_kwargs": {},
24
  "_format_type": null,
examples/AQA/WavCaps-QA-Test/dataset_info.json CHANGED
@@ -171,6 +171,42 @@
171
  "dtype": "string",
172
  "_type": "Value"
173
  }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
174
  }
175
  },
176
  "homepage": "",
 
171
  "dtype": "string",
172
  "_type": "Value"
173
  }
174
+ },
175
+ "meralion_audiollm_v1_mse": {
176
+ "answer": {
177
+ "dtype": "string",
178
+ "_type": "Value"
179
+ },
180
+ "model_prediction": {
181
+ "dtype": "string",
182
+ "_type": "Value"
183
+ },
184
+ "task_type": {
185
+ "dtype": "string",
186
+ "_type": "Value"
187
+ },
188
+ "text": {
189
+ "dtype": "string",
190
+ "_type": "Value"
191
+ }
192
+ },
193
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora": {
194
+ "answer": {
195
+ "dtype": "string",
196
+ "_type": "Value"
197
+ },
198
+ "model_prediction": {
199
+ "dtype": "string",
200
+ "_type": "Value"
201
+ },
202
+ "task_type": {
203
+ "dtype": "string",
204
+ "_type": "Value"
205
+ },
206
+ "text": {
207
+ "dtype": "string",
208
+ "_type": "Value"
209
+ }
210
  }
211
  },
212
  "homepage": "",
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": "2b00ba42b5d66bed",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
@@ -16,7 +16,9 @@
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
  "mowe_audio",
19
- "qwen_audio_chat"
 
 
20
  ],
21
  "_format_kwargs": {},
22
  "_format_type": null,
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "5aa9e7145a7a79a7",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
 
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
  "mowe_audio",
19
+ "qwen_audio_chat",
20
+ "meralion_audiollm_v1_mse",
21
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora"
22
  ],
23
  "_format_kwargs": {},
24
  "_format_type": null,
examples/AR/VoxCeleb-Accent-Test/dataset_info.json CHANGED
@@ -179,6 +179,42 @@
179
  "dtype": "string",
180
  "_type": "Value"
181
  }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
182
  }
183
  },
184
  "homepage": "",
 
179
  "dtype": "string",
180
  "_type": "Value"
181
  }
182
+ },
183
+ "meralion_audiollm_v1_mse": {
184
+ "answer": {
185
+ "dtype": "string",
186
+ "_type": "Value"
187
+ },
188
+ "model_prediction": {
189
+ "dtype": "string",
190
+ "_type": "Value"
191
+ },
192
+ "task_type": {
193
+ "dtype": "string",
194
+ "_type": "Value"
195
+ },
196
+ "text": {
197
+ "dtype": "string",
198
+ "_type": "Value"
199
+ }
200
+ },
201
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora": {
202
+ "answer": {
203
+ "dtype": "string",
204
+ "_type": "Value"
205
+ },
206
+ "model_prediction": {
207
+ "dtype": "string",
208
+ "_type": "Value"
209
+ },
210
+ "task_type": {
211
+ "dtype": "string",
212
+ "_type": "Value"
213
+ },
214
+ "text": {
215
+ "dtype": "string",
216
+ "_type": "Value"
217
+ }
218
  }
219
  },
220
  "homepage": "",
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": "8e8e0515e988a016",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
@@ -16,7 +16,9 @@
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
  "mowe_audio",
19
- "qwen_audio_chat"
 
 
20
  ],
21
  "_format_kwargs": {},
22
  "_format_type": null,
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "196da8d0b0da0eb9",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
 
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
  "mowe_audio",
19
+ "qwen_audio_chat",
20
+ "meralion_audiollm_v1_mse",
21
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora"
22
  ],
23
  "_format_kwargs": {},
24
  "_format_type": null,
examples/ASR/Common-Voice-15-En-Test/dataset_info.json CHANGED
@@ -38,7 +38,7 @@
38
  "_type": "Value"
39
  },
40
  "age": {
41
- "dtype": "null",
42
  "_type": "Value"
43
  },
44
  "client_id": {
@@ -199,6 +199,42 @@
199
  "dtype": "string",
200
  "_type": "Value"
201
  }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
202
  }
203
  },
204
  "homepage": "",
 
38
  "_type": "Value"
39
  },
40
  "age": {
41
+ "dtype": "string",
42
  "_type": "Value"
43
  },
44
  "client_id": {
 
199
  "dtype": "string",
200
  "_type": "Value"
201
  }
202
+ },
203
+ "meralion_audiollm_v1_mse": {
204
+ "answer": {
205
+ "dtype": "string",
206
+ "_type": "Value"
207
+ },
208
+ "model_prediction": {
209
+ "dtype": "string",
210
+ "_type": "Value"
211
+ },
212
+ "task_type": {
213
+ "dtype": "string",
214
+ "_type": "Value"
215
+ },
216
+ "text": {
217
+ "dtype": "string",
218
+ "_type": "Value"
219
+ }
220
+ },
221
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora": {
222
+ "answer": {
223
+ "dtype": "string",
224
+ "_type": "Value"
225
+ },
226
+ "model_prediction": {
227
+ "dtype": "string",
228
+ "_type": "Value"
229
+ },
230
+ "task_type": {
231
+ "dtype": "string",
232
+ "_type": "Value"
233
+ },
234
+ "text": {
235
+ "dtype": "string",
236
+ "_type": "Value"
237
+ }
238
  }
239
  },
240
  "homepage": "",
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": "23bec5037b5ce6a4",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
@@ -16,7 +16,9 @@
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
  "mowe_audio",
19
- "qwen_audio_chat"
 
 
20
  ],
21
  "_format_kwargs": {},
22
  "_format_type": null,
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "c8a3251745f6df7c",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
 
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
  "mowe_audio",
19
+ "qwen_audio_chat",
20
+ "meralion_audiollm_v1_mse",
21
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora"
22
  ],
23
  "_format_kwargs": {},
24
  "_format_type": null,
examples/ASR/Earnings21-Test/dataset_info.json CHANGED
@@ -163,6 +163,42 @@
163
  "dtype": "string",
164
  "_type": "Value"
165
  }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
166
  }
167
  },
168
  "homepage": "",
 
163
  "dtype": "string",
164
  "_type": "Value"
165
  }
166
+ },
167
+ "meralion_audiollm_v1_mse": {
168
+ "answer": {
169
+ "dtype": "string",
170
+ "_type": "Value"
171
+ },
172
+ "model_prediction": {
173
+ "dtype": "string",
174
+ "_type": "Value"
175
+ },
176
+ "task_type": {
177
+ "dtype": "string",
178
+ "_type": "Value"
179
+ },
180
+ "text": {
181
+ "dtype": "string",
182
+ "_type": "Value"
183
+ }
184
+ },
185
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora": {
186
+ "answer": {
187
+ "dtype": "string",
188
+ "_type": "Value"
189
+ },
190
+ "model_prediction": {
191
+ "dtype": "string",
192
+ "_type": "Value"
193
+ },
194
+ "task_type": {
195
+ "dtype": "string",
196
+ "_type": "Value"
197
+ },
198
+ "text": {
199
+ "dtype": "string",
200
+ "_type": "Value"
201
+ }
202
  }
203
  },
204
  "homepage": "",
examples/ASR/Earnings21-Test/state.json CHANGED
@@ -4,7 +4,7 @@
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
- "_fingerprint": "6d8e11ac5a63a2d2",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
@@ -16,7 +16,9 @@
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
  "mowe_audio",
19
- "qwen_audio_chat"
 
 
20
  ],
21
  "_format_kwargs": {},
22
  "_format_type": null,
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "fee62a469e646020",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
 
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
  "mowe_audio",
19
+ "qwen_audio_chat",
20
+ "meralion_audiollm_v1_mse",
21
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora"
22
  ],
23
  "_format_kwargs": {},
24
  "_format_type": null,
examples/ASR/Earnings22-Test/dataset_info.json CHANGED
@@ -92,6 +92,24 @@
92
  "_type": "Value"
93
  }
94
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
95
  "whisper_large_v3_with_llama_3_8b_instruct": {
96
  "answer": {
97
  "dtype": "string",
@@ -145,6 +163,42 @@
145
  "dtype": "string",
146
  "_type": "Value"
147
  }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
148
  }
149
  },
150
  "homepage": "",
 
92
  "_type": "Value"
93
  }
94
  },
95
+ "meralion_audiollm_v1_lora": {
96
+ "answer": {
97
+ "dtype": "string",
98
+ "_type": "Value"
99
+ },
100
+ "model_prediction": {
101
+ "dtype": "string",
102
+ "_type": "Value"
103
+ },
104
+ "task_type": {
105
+ "dtype": "string",
106
+ "_type": "Value"
107
+ },
108
+ "text": {
109
+ "dtype": "string",
110
+ "_type": "Value"
111
+ }
112
+ },
113
  "whisper_large_v3_with_llama_3_8b_instruct": {
114
  "answer": {
115
  "dtype": "string",
 
163
  "dtype": "string",
164
  "_type": "Value"
165
  }
166
+ },
167
+ "meralion_audiollm_v1_mse": {
168
+ "answer": {
169
+ "dtype": "string",
170
+ "_type": "Value"
171
+ },
172
+ "model_prediction": {
173
+ "dtype": "string",
174
+ "_type": "Value"
175
+ },
176
+ "task_type": {
177
+ "dtype": "string",
178
+ "_type": "Value"
179
+ },
180
+ "text": {
181
+ "dtype": "string",
182
+ "_type": "Value"
183
+ }
184
+ },
185
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora": {
186
+ "answer": {
187
+ "dtype": "string",
188
+ "_type": "Value"
189
+ },
190
+ "model_prediction": {
191
+ "dtype": "string",
192
+ "_type": "Value"
193
+ },
194
+ "task_type": {
195
+ "dtype": "string",
196
+ "_type": "Value"
197
+ },
198
+ "text": {
199
+ "dtype": "string",
200
+ "_type": "Value"
201
+ }
202
  }
203
  },
204
  "homepage": "",
examples/ASR/Earnings22-Test/state.json CHANGED
@@ -4,7 +4,7 @@
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
- "_fingerprint": "f71b90ac0caefff8",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
@@ -13,9 +13,12 @@
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,
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "c7f7dfbe71f40f5a",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
 
13
  "salmonn_7b",
14
  "wavllm_fairseq",
15
  "Qwen2-Audio-7B-Instruct",
16
+ "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
  "mowe_audio",
19
+ "qwen_audio_chat",
20
+ "meralion_audiollm_v1_mse",
21
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora"
22
  ],
23
  "_format_kwargs": {},
24
  "_format_type": null,
examples/ASR/GigaSpeech-Test/dataset_info.json CHANGED
@@ -191,6 +191,42 @@
191
  "dtype": "string",
192
  "_type": "Value"
193
  }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
194
  }
195
  },
196
  "homepage": "",
 
191
  "dtype": "string",
192
  "_type": "Value"
193
  }
194
+ },
195
+ "meralion_audiollm_v1_mse": {
196
+ "answer": {
197
+ "dtype": "string",
198
+ "_type": "Value"
199
+ },
200
+ "model_prediction": {
201
+ "dtype": "string",
202
+ "_type": "Value"
203
+ },
204
+ "task_type": {
205
+ "dtype": "string",
206
+ "_type": "Value"
207
+ },
208
+ "text": {
209
+ "dtype": "string",
210
+ "_type": "Value"
211
+ }
212
+ },
213
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora": {
214
+ "answer": {
215
+ "dtype": "string",
216
+ "_type": "Value"
217
+ },
218
+ "model_prediction": {
219
+ "dtype": "string",
220
+ "_type": "Value"
221
+ },
222
+ "task_type": {
223
+ "dtype": "string",
224
+ "_type": "Value"
225
+ },
226
+ "text": {
227
+ "dtype": "string",
228
+ "_type": "Value"
229
+ }
230
  }
231
  },
232
  "homepage": "",
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": "67d1ab1b99556a9f",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
@@ -16,7 +16,9 @@
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
  "mowe_audio",
19
- "qwen_audio_chat"
 
 
20
  ],
21
  "_format_kwargs": {},
22
  "_format_type": null,
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "749a511e2cc30275",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
 
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
  "mowe_audio",
19
+ "qwen_audio_chat",
20
+ "meralion_audiollm_v1_mse",
21
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora"
22
  ],
23
  "_format_kwargs": {},
24
  "_format_type": null,
examples/ASR/IMDA-Part1-ASR-Test/dataset_info.json CHANGED
@@ -193,6 +193,42 @@
193
  "dtype": "string",
194
  "_type": "Value"
195
  }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
196
  }
197
  },
198
  "homepage": "",
 
193
  "dtype": "string",
194
  "_type": "Value"
195
  }
196
+ },
197
+ "meralion_audiollm_v1_mse": {
198
+ "answer": {
199
+ "dtype": "string",
200
+ "_type": "Value"
201
+ },
202
+ "model_prediction": {
203
+ "dtype": "string",
204
+ "_type": "Value"
205
+ },
206
+ "task_type": {
207
+ "dtype": "string",
208
+ "_type": "Value"
209
+ },
210
+ "text": {
211
+ "dtype": "string",
212
+ "_type": "Value"
213
+ }
214
+ },
215
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora": {
216
+ "answer": {
217
+ "dtype": "string",
218
+ "_type": "Value"
219
+ },
220
+ "model_prediction": {
221
+ "dtype": "string",
222
+ "_type": "Value"
223
+ },
224
+ "task_type": {
225
+ "dtype": "string",
226
+ "_type": "Value"
227
+ },
228
+ "text": {
229
+ "dtype": "string",
230
+ "_type": "Value"
231
+ }
232
  }
233
  },
234
  "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
@@ -4,7 +4,7 @@
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
- "_fingerprint": "6de71e0f4c76af43",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
@@ -15,7 +15,9 @@
15
  "Qwen2-Audio-7B-Instruct",
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
- "qwen_audio_chat"
 
 
19
  ],
20
  "_format_kwargs": {},
21
  "_format_type": null,
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "9419619e05bfc9dc",
8
  "_format_columns": [
9
  "context",
10
  "instruction",
 
15
  "Qwen2-Audio-7B-Instruct",
16
  "meralion_audiollm_v1_lora",
17
  "whisper_large_v3_with_llama_3_8b_instruct",
18
+ "qwen_audio_chat",
19
+ "meralion_audiollm_v1_mse",
20
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora"
21
  ],
22
  "_format_kwargs": {},
23
  "_format_type": null,
examples/ASR/IMDA-Part2-ASR-Test/dataset_info.json CHANGED
@@ -193,6 +193,42 @@
193
  "dtype": "string",
194
  "_type": "Value"
195
  }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
196
  }
197
  },
198
  "homepage": "",
 
193
  "dtype": "string",
194
  "_type": "Value"
195
  }
196
+ },
197
+ "meralion_audiollm_v1_mse": {
198
+ "answer": {
199
+ "dtype": "string",
200
+ "_type": "Value"
201
+ },
202
+ "model_prediction": {
203
+ "dtype": "string",
204
+ "_type": "Value"
205
+ },
206
+ "task_type": {
207
+ "dtype": "string",
208
+ "_type": "Value"
209
+ },
210
+ "text": {
211
+ "dtype": "string",
212
+ "_type": "Value"
213
+ }
214
+ },
215
+ "stage2_whisper3_fft_mlp100_gemma2_9b_lora": {
216
+ "answer": {
217
+ "dtype": "string",
218
+ "_type": "Value"
219
+ },
220
+ "model_prediction": {
221
+ "dtype": "string",
222
+ "_type": "Value"
223
+ },
224
+ "task_type": {
225
+ "dtype": "string",
226
+ "_type": "Value"
227
+ },
228
+ "text": {
229
+ "dtype": "string",
230
+ "_type": "Value"
231
+ }
232
  }
233
  },
234
  "homepage": "",