Upload benchmark_results.json with huggingface_hub
Browse files- benchmark_results.json +77 -77
benchmark_results.json
CHANGED
|
@@ -1,82 +1,82 @@
|
|
| 1 |
{
|
| 2 |
-
"model": "/mnt/aynur/
|
| 3 |
-
"timestamp": "2025-11-
|
| 4 |
-
"overall_accuracy":
|
| 5 |
"results": {
|
| 6 |
"Pure Music": {
|
| 7 |
"accuracy": 100.0,
|
| 8 |
"correct": 10,
|
| 9 |
"total": 10,
|
| 10 |
-
"avg_music_score": 0.
|
| 11 |
-
"avg_speech_score":
|
| 12 |
"files": [
|
| 13 |
{
|
| 14 |
"file": "music_1.wav",
|
| 15 |
-
"music": 0.
|
| 16 |
-
"speech":
|
| 17 |
"predicted": "music",
|
| 18 |
"correct": true
|
| 19 |
},
|
| 20 |
{
|
| 21 |
"file": "music_10.wav",
|
| 22 |
-
"music": 0.
|
| 23 |
-
"speech":
|
| 24 |
"predicted": "music",
|
| 25 |
"correct": true
|
| 26 |
},
|
| 27 |
{
|
| 28 |
"file": "music_2.wav",
|
| 29 |
-
"music": 0.
|
| 30 |
-
"speech":
|
| 31 |
"predicted": "music",
|
| 32 |
"correct": true
|
| 33 |
},
|
| 34 |
{
|
| 35 |
"file": "music_3.wav",
|
| 36 |
-
"music": 0.
|
| 37 |
-
"speech":
|
| 38 |
"predicted": "music",
|
| 39 |
"correct": true
|
| 40 |
},
|
| 41 |
{
|
| 42 |
"file": "music_4.wav",
|
| 43 |
-
"music": 0.
|
| 44 |
-
"speech":
|
| 45 |
"predicted": "music",
|
| 46 |
"correct": true
|
| 47 |
},
|
| 48 |
{
|
| 49 |
"file": "music_5.wav",
|
| 50 |
-
"music": 0.
|
| 51 |
-
"speech":
|
| 52 |
"predicted": "music",
|
| 53 |
"correct": true
|
| 54 |
},
|
| 55 |
{
|
| 56 |
"file": "music_6.wav",
|
| 57 |
-
"music": 0.
|
| 58 |
-
"speech":
|
| 59 |
"predicted": "music",
|
| 60 |
"correct": true
|
| 61 |
},
|
| 62 |
{
|
| 63 |
"file": "music_7.wav",
|
| 64 |
-
"music": 0.
|
| 65 |
-
"speech":
|
| 66 |
"predicted": "music",
|
| 67 |
"correct": true
|
| 68 |
},
|
| 69 |
{
|
| 70 |
"file": "music_8.wav",
|
| 71 |
-
"music": 0.
|
| 72 |
-
"speech":
|
| 73 |
"predicted": "music",
|
| 74 |
"correct": true
|
| 75 |
},
|
| 76 |
{
|
| 77 |
"file": "music_9.wav",
|
| 78 |
-
"music": 0.
|
| 79 |
-
"speech":
|
| 80 |
"predicted": "music",
|
| 81 |
"correct": true
|
| 82 |
}
|
|
@@ -86,155 +86,155 @@
|
|
| 86 |
"accuracy": 70.0,
|
| 87 |
"correct": 7,
|
| 88 |
"total": 10,
|
| 89 |
-
"avg_music_score": 0.
|
| 90 |
-
"avg_speech_score": 0.
|
| 91 |
"files": [
|
| 92 |
{
|
| 93 |
"file": "speech_1.wav",
|
| 94 |
-
"music":
|
| 95 |
-
"speech": 0.
|
| 96 |
"predicted": "speech",
|
| 97 |
"correct": true
|
| 98 |
},
|
| 99 |
{
|
| 100 |
"file": "speech_10.wav",
|
| 101 |
-
"music":
|
| 102 |
-
"speech": 0.
|
| 103 |
"predicted": "speech",
|
| 104 |
"correct": true
|
| 105 |
},
|
| 106 |
{
|
| 107 |
"file": "speech_2.wav",
|
| 108 |
-
"music":
|
| 109 |
-
"speech": 0.
|
| 110 |
"predicted": "speech",
|
| 111 |
"correct": true
|
| 112 |
},
|
| 113 |
{
|
| 114 |
"file": "speech_3.wav",
|
| 115 |
-
"music": 0.
|
| 116 |
-
"speech": 0.
|
| 117 |
"predicted": "music",
|
| 118 |
"correct": false
|
| 119 |
},
|
| 120 |
{
|
| 121 |
"file": "speech_4.wav",
|
| 122 |
-
"music": 0.
|
| 123 |
-
"speech": 0.
|
| 124 |
"predicted": "music",
|
| 125 |
"correct": false
|
| 126 |
},
|
| 127 |
{
|
| 128 |
"file": "speech_5.wav",
|
| 129 |
-
"music": 0.
|
| 130 |
-
"speech":
|
| 131 |
-
"predicted": "
|
| 132 |
-
"correct":
|
| 133 |
},
|
| 134 |
{
|
| 135 |
"file": "speech_6.wav",
|
| 136 |
-
"music": 0.
|
| 137 |
-
"speech": 0.
|
| 138 |
-
"predicted": "
|
| 139 |
-
"correct":
|
| 140 |
},
|
| 141 |
{
|
| 142 |
"file": "speech_7.wav",
|
| 143 |
-
"music": 0.
|
| 144 |
-
"speech": 0.
|
| 145 |
"predicted": "speech",
|
| 146 |
"correct": true
|
| 147 |
},
|
| 148 |
{
|
| 149 |
"file": "speech_8.wav",
|
| 150 |
-
"music": 0.
|
| 151 |
-
"speech": 0.
|
| 152 |
"predicted": "speech",
|
| 153 |
"correct": true
|
| 154 |
},
|
| 155 |
{
|
| 156 |
"file": "speech_9.wav",
|
| 157 |
-
"music":
|
| 158 |
-
"speech": 0.
|
| 159 |
"predicted": "speech",
|
| 160 |
"correct": true
|
| 161 |
}
|
| 162 |
]
|
| 163 |
},
|
| 164 |
"Speech + Music": {
|
| 165 |
-
"accuracy":
|
| 166 |
-
"correct":
|
| 167 |
"total": 10,
|
| 168 |
-
"avg_music_score": 0.
|
| 169 |
-
"avg_speech_score": 0.
|
| 170 |
"files": [
|
| 171 |
{
|
| 172 |
"file": "speech_and_music_1.wav",
|
| 173 |
-
"music": 0.
|
| 174 |
-
"speech":
|
| 175 |
"predicted": "music",
|
| 176 |
"correct": true
|
| 177 |
},
|
| 178 |
{
|
| 179 |
"file": "speech_and_music_10.wav",
|
| 180 |
-
"music": 0.
|
| 181 |
-
"speech":
|
| 182 |
"predicted": "music",
|
| 183 |
"correct": true
|
| 184 |
},
|
| 185 |
{
|
| 186 |
"file": "speech_and_music_2.wav",
|
| 187 |
-
"music": 0.
|
| 188 |
-
"speech":
|
| 189 |
"predicted": "music",
|
| 190 |
"correct": true
|
| 191 |
},
|
| 192 |
{
|
| 193 |
"file": "speech_and_music_3wav.wav",
|
| 194 |
-
"music": 0.
|
| 195 |
-
"speech":
|
| 196 |
"predicted": "music",
|
| 197 |
"correct": true
|
| 198 |
},
|
| 199 |
{
|
| 200 |
"file": "speech_and_music_4.wav",
|
| 201 |
-
"music": 0.
|
| 202 |
-
"speech":
|
| 203 |
"predicted": "music",
|
| 204 |
"correct": true
|
| 205 |
},
|
| 206 |
{
|
| 207 |
"file": "speech_and_music_5.wav",
|
| 208 |
-
"music": 0.
|
| 209 |
-
"speech":
|
| 210 |
"predicted": "music",
|
| 211 |
"correct": true
|
| 212 |
},
|
| 213 |
{
|
| 214 |
"file": "speech_and_music_6.wav",
|
| 215 |
-
"music": 0.
|
| 216 |
-
"speech":
|
| 217 |
"predicted": "music",
|
| 218 |
"correct": true
|
| 219 |
},
|
| 220 |
{
|
| 221 |
"file": "speech_and_music_7.wav",
|
| 222 |
-
"music": 0.
|
| 223 |
-
"speech":
|
| 224 |
-
"predicted": "
|
| 225 |
-
"correct":
|
| 226 |
},
|
| 227 |
{
|
| 228 |
"file": "speech_and_music_8.wav",
|
| 229 |
-
"music": 0.
|
| 230 |
-
"speech":
|
| 231 |
"predicted": "music",
|
| 232 |
"correct": true
|
| 233 |
},
|
| 234 |
{
|
| 235 |
"file": "speech_and_music_9.wav",
|
| 236 |
-
"music": 0.
|
| 237 |
-
"speech":
|
| 238 |
"predicted": "music",
|
| 239 |
"correct": true
|
| 240 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"model": "/mnt/aynur/ast_music_1k_1epoch/final_model",
|
| 3 |
+
"timestamp": "2025-11-26T15:28:56.040313",
|
| 4 |
+
"overall_accuracy": 90.0,
|
| 5 |
"results": {
|
| 6 |
"Pure Music": {
|
| 7 |
"accuracy": 100.0,
|
| 8 |
"correct": 10,
|
| 9 |
"total": 10,
|
| 10 |
+
"avg_music_score": 0.999993908405304,
|
| 11 |
+
"avg_speech_score": 6.091594696044922e-06,
|
| 12 |
"files": [
|
| 13 |
{
|
| 14 |
"file": "music_1.wav",
|
| 15 |
+
"music": 0.9999896287918091,
|
| 16 |
+
"speech": 1.0371208190917969e-05,
|
| 17 |
"predicted": "music",
|
| 18 |
"correct": true
|
| 19 |
},
|
| 20 |
{
|
| 21 |
"file": "music_10.wav",
|
| 22 |
+
"music": 0.999994158744812,
|
| 23 |
+
"speech": 5.841255187988281e-06,
|
| 24 |
"predicted": "music",
|
| 25 |
"correct": true
|
| 26 |
},
|
| 27 |
{
|
| 28 |
"file": "music_2.wav",
|
| 29 |
+
"music": 0.999995231628418,
|
| 30 |
+
"speech": 4.76837158203125e-06,
|
| 31 |
"predicted": "music",
|
| 32 |
"correct": true
|
| 33 |
},
|
| 34 |
{
|
| 35 |
"file": "music_3.wav",
|
| 36 |
+
"music": 0.9999922513961792,
|
| 37 |
+
"speech": 7.748603820800781e-06,
|
| 38 |
"predicted": "music",
|
| 39 |
"correct": true
|
| 40 |
},
|
| 41 |
{
|
| 42 |
"file": "music_4.wav",
|
| 43 |
+
"music": 0.9999986886978149,
|
| 44 |
+
"speech": 1.3113021850585938e-06,
|
| 45 |
"predicted": "music",
|
| 46 |
"correct": true
|
| 47 |
},
|
| 48 |
{
|
| 49 |
"file": "music_5.wav",
|
| 50 |
+
"music": 0.9999957084655762,
|
| 51 |
+
"speech": 4.291534423828125e-06,
|
| 52 |
"predicted": "music",
|
| 53 |
"correct": true
|
| 54 |
},
|
| 55 |
{
|
| 56 |
"file": "music_6.wav",
|
| 57 |
+
"music": 0.9999951124191284,
|
| 58 |
+
"speech": 4.887580871582031e-06,
|
| 59 |
"predicted": "music",
|
| 60 |
"correct": true
|
| 61 |
},
|
| 62 |
{
|
| 63 |
"file": "music_7.wav",
|
| 64 |
+
"music": 0.9999871253967285,
|
| 65 |
+
"speech": 1.2874603271484375e-05,
|
| 66 |
"predicted": "music",
|
| 67 |
"correct": true
|
| 68 |
},
|
| 69 |
{
|
| 70 |
"file": "music_8.wav",
|
| 71 |
+
"music": 0.9999980926513672,
|
| 72 |
+
"speech": 1.9073486328125e-06,
|
| 73 |
"predicted": "music",
|
| 74 |
"correct": true
|
| 75 |
},
|
| 76 |
{
|
| 77 |
"file": "music_9.wav",
|
| 78 |
+
"music": 0.999993085861206,
|
| 79 |
+
"speech": 6.9141387939453125e-06,
|
| 80 |
"predicted": "music",
|
| 81 |
"correct": true
|
| 82 |
}
|
|
|
|
| 86 |
"accuracy": 70.0,
|
| 87 |
"correct": 7,
|
| 88 |
"total": 10,
|
| 89 |
+
"avg_music_score": 0.3116513199704059,
|
| 90 |
+
"avg_speech_score": 0.6883486800295942,
|
| 91 |
"files": [
|
| 92 |
{
|
| 93 |
"file": "speech_1.wav",
|
| 94 |
+
"music": 0.0002677915617823601,
|
| 95 |
+
"speech": 0.9997322084382176,
|
| 96 |
"predicted": "speech",
|
| 97 |
"correct": true
|
| 98 |
},
|
| 99 |
{
|
| 100 |
"file": "speech_10.wav",
|
| 101 |
+
"music": 0.0022506339009851217,
|
| 102 |
+
"speech": 0.9977493660990149,
|
| 103 |
"predicted": "speech",
|
| 104 |
"correct": true
|
| 105 |
},
|
| 106 |
{
|
| 107 |
"file": "speech_2.wav",
|
| 108 |
+
"music": 0.0001184314678539522,
|
| 109 |
+
"speech": 0.999881568532146,
|
| 110 |
"predicted": "speech",
|
| 111 |
"correct": true
|
| 112 |
},
|
| 113 |
{
|
| 114 |
"file": "speech_3.wav",
|
| 115 |
+
"music": 0.7140341401100159,
|
| 116 |
+
"speech": 0.28596585988998413,
|
| 117 |
"predicted": "music",
|
| 118 |
"correct": false
|
| 119 |
},
|
| 120 |
{
|
| 121 |
"file": "speech_4.wav",
|
| 122 |
+
"music": 0.9064487218856812,
|
| 123 |
+
"speech": 0.09355127811431885,
|
| 124 |
"predicted": "music",
|
| 125 |
"correct": false
|
| 126 |
},
|
| 127 |
{
|
| 128 |
"file": "speech_5.wav",
|
| 129 |
+
"music": 0.35047778487205505,
|
| 130 |
+
"speech": 0.649522215127945,
|
| 131 |
+
"predicted": "speech",
|
| 132 |
+
"correct": true
|
| 133 |
},
|
| 134 |
{
|
| 135 |
"file": "speech_6.wav",
|
| 136 |
+
"music": 0.895337700843811,
|
| 137 |
+
"speech": 0.10466229915618896,
|
| 138 |
+
"predicted": "music",
|
| 139 |
+
"correct": false
|
| 140 |
},
|
| 141 |
{
|
| 142 |
"file": "speech_7.wav",
|
| 143 |
+
"music": 0.0676170364022255,
|
| 144 |
+
"speech": 0.9323829635977745,
|
| 145 |
"predicted": "speech",
|
| 146 |
"correct": true
|
| 147 |
},
|
| 148 |
{
|
| 149 |
"file": "speech_8.wav",
|
| 150 |
+
"music": 0.0965794250369072,
|
| 151 |
+
"speech": 0.9034205749630928,
|
| 152 |
"predicted": "speech",
|
| 153 |
"correct": true
|
| 154 |
},
|
| 155 |
{
|
| 156 |
"file": "speech_9.wav",
|
| 157 |
+
"music": 0.0833815336227417,
|
| 158 |
+
"speech": 0.9166184663772583,
|
| 159 |
"predicted": "speech",
|
| 160 |
"correct": true
|
| 161 |
}
|
| 162 |
]
|
| 163 |
},
|
| 164 |
"Speech + Music": {
|
| 165 |
+
"accuracy": 100.0,
|
| 166 |
+
"correct": 10,
|
| 167 |
"total": 10,
|
| 168 |
+
"avg_music_score": 0.9979188024997712,
|
| 169 |
+
"avg_speech_score": 0.0020811975002288817,
|
| 170 |
"files": [
|
| 171 |
{
|
| 172 |
"file": "speech_and_music_1.wav",
|
| 173 |
+
"music": 0.9948858618736267,
|
| 174 |
+
"speech": 0.005114138126373291,
|
| 175 |
"predicted": "music",
|
| 176 |
"correct": true
|
| 177 |
},
|
| 178 |
{
|
| 179 |
"file": "speech_and_music_10.wav",
|
| 180 |
+
"music": 0.9871156811714172,
|
| 181 |
+
"speech": 0.012884318828582764,
|
| 182 |
"predicted": "music",
|
| 183 |
"correct": true
|
| 184 |
},
|
| 185 |
{
|
| 186 |
"file": "speech_and_music_2.wav",
|
| 187 |
+
"music": 0.9998169541358948,
|
| 188 |
+
"speech": 0.0001830458641052246,
|
| 189 |
"predicted": "music",
|
| 190 |
"correct": true
|
| 191 |
},
|
| 192 |
{
|
| 193 |
"file": "speech_and_music_3wav.wav",
|
| 194 |
+
"music": 0.9997709393501282,
|
| 195 |
+
"speech": 0.00022906064987182617,
|
| 196 |
"predicted": "music",
|
| 197 |
"correct": true
|
| 198 |
},
|
| 199 |
{
|
| 200 |
"file": "speech_and_music_4.wav",
|
| 201 |
+
"music": 0.9999324083328247,
|
| 202 |
+
"speech": 6.759166717529297e-05,
|
| 203 |
"predicted": "music",
|
| 204 |
"correct": true
|
| 205 |
},
|
| 206 |
{
|
| 207 |
"file": "speech_and_music_5.wav",
|
| 208 |
+
"music": 0.9982500672340393,
|
| 209 |
+
"speech": 0.0017499327659606934,
|
| 210 |
"predicted": "music",
|
| 211 |
"correct": true
|
| 212 |
},
|
| 213 |
{
|
| 214 |
"file": "speech_and_music_6.wav",
|
| 215 |
+
"music": 0.9998733997344971,
|
| 216 |
+
"speech": 0.0001266002655029297,
|
| 217 |
"predicted": "music",
|
| 218 |
"correct": true
|
| 219 |
},
|
| 220 |
{
|
| 221 |
"file": "speech_and_music_7.wav",
|
| 222 |
+
"music": 0.9999369382858276,
|
| 223 |
+
"speech": 6.306171417236328e-05,
|
| 224 |
+
"predicted": "music",
|
| 225 |
+
"correct": true
|
| 226 |
},
|
| 227 |
{
|
| 228 |
"file": "speech_and_music_8.wav",
|
| 229 |
+
"music": 0.9996254444122314,
|
| 230 |
+
"speech": 0.0003745555877685547,
|
| 231 |
"predicted": "music",
|
| 232 |
"correct": true
|
| 233 |
},
|
| 234 |
{
|
| 235 |
"file": "speech_and_music_9.wav",
|
| 236 |
+
"music": 0.9999803304672241,
|
| 237 |
+
"speech": 1.9669532775878906e-05,
|
| 238 |
"predicted": "music",
|
| 239 |
"correct": true
|
| 240 |
}
|