Upload benchmark_results.json with huggingface_hub
Browse files- benchmark_results.json +244 -0
benchmark_results.json
ADDED
|
@@ -0,0 +1,244 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model": "./ast_music_45k_1epoch/final_model",
|
| 3 |
+
"timestamp": "2025-11-26T15:47:30.892279",
|
| 4 |
+
"overall_accuracy": 86.66666666666667,
|
| 5 |
+
"results": {
|
| 6 |
+
"Pure Music": {
|
| 7 |
+
"accuracy": 100.0,
|
| 8 |
+
"correct": 10,
|
| 9 |
+
"total": 10,
|
| 10 |
+
"avg_music_score": 0.9989962816238404,
|
| 11 |
+
"avg_speech_score": 0.0010037183761596679,
|
| 12 |
+
"files": [
|
| 13 |
+
{
|
| 14 |
+
"file": "music_1.wav",
|
| 15 |
+
"music": 0.9999573230743408,
|
| 16 |
+
"speech": 4.267692565917969e-05,
|
| 17 |
+
"predicted": "music",
|
| 18 |
+
"correct": true
|
| 19 |
+
},
|
| 20 |
+
{
|
| 21 |
+
"file": "music_10.wav",
|
| 22 |
+
"music": 0.9988263249397278,
|
| 23 |
+
"speech": 0.0011736750602722168,
|
| 24 |
+
"predicted": "music",
|
| 25 |
+
"correct": true
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"file": "music_2.wav",
|
| 29 |
+
"music": 0.9991300702095032,
|
| 30 |
+
"speech": 0.0008699297904968262,
|
| 31 |
+
"predicted": "music",
|
| 32 |
+
"correct": true
|
| 33 |
+
},
|
| 34 |
+
{
|
| 35 |
+
"file": "music_3.wav",
|
| 36 |
+
"music": 0.9988117218017578,
|
| 37 |
+
"speech": 0.0011882781982421875,
|
| 38 |
+
"predicted": "music",
|
| 39 |
+
"correct": true
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"file": "music_4.wav",
|
| 43 |
+
"music": 0.9997615218162537,
|
| 44 |
+
"speech": 0.0002384781837463379,
|
| 45 |
+
"predicted": "music",
|
| 46 |
+
"correct": true
|
| 47 |
+
},
|
| 48 |
+
{
|
| 49 |
+
"file": "music_5.wav",
|
| 50 |
+
"music": 0.9955949187278748,
|
| 51 |
+
"speech": 0.004405081272125244,
|
| 52 |
+
"predicted": "music",
|
| 53 |
+
"correct": true
|
| 54 |
+
},
|
| 55 |
+
{
|
| 56 |
+
"file": "music_6.wav",
|
| 57 |
+
"music": 0.9998776912689209,
|
| 58 |
+
"speech": 0.00012230873107910156,
|
| 59 |
+
"predicted": "music",
|
| 60 |
+
"correct": true
|
| 61 |
+
},
|
| 62 |
+
{
|
| 63 |
+
"file": "music_7.wav",
|
| 64 |
+
"music": 0.9981030225753784,
|
| 65 |
+
"speech": 0.001896977424621582,
|
| 66 |
+
"predicted": "music",
|
| 67 |
+
"correct": true
|
| 68 |
+
},
|
| 69 |
+
{
|
| 70 |
+
"file": "music_8.wav",
|
| 71 |
+
"music": 0.9999755620956421,
|
| 72 |
+
"speech": 2.4437904357910156e-05,
|
| 73 |
+
"predicted": "music",
|
| 74 |
+
"correct": true
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"file": "music_9.wav",
|
| 78 |
+
"music": 0.9999246597290039,
|
| 79 |
+
"speech": 7.534027099609375e-05,
|
| 80 |
+
"predicted": "music",
|
| 81 |
+
"correct": true
|
| 82 |
+
}
|
| 83 |
+
]
|
| 84 |
+
},
|
| 85 |
+
"Pure Speech": {
|
| 86 |
+
"accuracy": 70.0,
|
| 87 |
+
"correct": 7,
|
| 88 |
+
"total": 10,
|
| 89 |
+
"avg_music_score": 0.2843911386182299,
|
| 90 |
+
"avg_speech_score": 0.7156088613817702,
|
| 91 |
+
"files": [
|
| 92 |
+
{
|
| 93 |
+
"file": "speech_1.wav",
|
| 94 |
+
"music": 7.125085539883003e-05,
|
| 95 |
+
"speech": 0.9999287491446012,
|
| 96 |
+
"predicted": "speech",
|
| 97 |
+
"correct": true
|
| 98 |
+
},
|
| 99 |
+
{
|
| 100 |
+
"file": "speech_10.wav",
|
| 101 |
+
"music": 9.815816883929074e-05,
|
| 102 |
+
"speech": 0.9999018418311607,
|
| 103 |
+
"predicted": "speech",
|
| 104 |
+
"correct": true
|
| 105 |
+
},
|
| 106 |
+
{
|
| 107 |
+
"file": "speech_2.wav",
|
| 108 |
+
"music": 5.82359243708197e-05,
|
| 109 |
+
"speech": 0.9999417640756292,
|
| 110 |
+
"predicted": "speech",
|
| 111 |
+
"correct": true
|
| 112 |
+
},
|
| 113 |
+
{
|
| 114 |
+
"file": "speech_3.wav",
|
| 115 |
+
"music": 0.823881983757019,
|
| 116 |
+
"speech": 0.17611801624298096,
|
| 117 |
+
"predicted": "music",
|
| 118 |
+
"correct": false
|
| 119 |
+
},
|
| 120 |
+
{
|
| 121 |
+
"file": "speech_4.wav",
|
| 122 |
+
"music": 0.9778100252151489,
|
| 123 |
+
"speech": 0.022189974784851074,
|
| 124 |
+
"predicted": "music",
|
| 125 |
+
"correct": false
|
| 126 |
+
},
|
| 127 |
+
{
|
| 128 |
+
"file": "speech_5.wav",
|
| 129 |
+
"music": 0.9999781847000122,
|
| 130 |
+
"speech": 2.181529998779297e-05,
|
| 131 |
+
"predicted": "music",
|
| 132 |
+
"correct": false
|
| 133 |
+
},
|
| 134 |
+
{
|
| 135 |
+
"file": "speech_6.wav",
|
| 136 |
+
"music": 0.03774089366197586,
|
| 137 |
+
"speech": 0.9622591063380241,
|
| 138 |
+
"predicted": "speech",
|
| 139 |
+
"correct": true
|
| 140 |
+
},
|
| 141 |
+
{
|
| 142 |
+
"file": "speech_7.wav",
|
| 143 |
+
"music": 0.0032406761310994625,
|
| 144 |
+
"speech": 0.9967593238689005,
|
| 145 |
+
"predicted": "speech",
|
| 146 |
+
"correct": true
|
| 147 |
+
},
|
| 148 |
+
{
|
| 149 |
+
"file": "speech_8.wav",
|
| 150 |
+
"music": 0.0009782465640455484,
|
| 151 |
+
"speech": 0.9990217534359545,
|
| 152 |
+
"predicted": "speech",
|
| 153 |
+
"correct": true
|
| 154 |
+
},
|
| 155 |
+
{
|
| 156 |
+
"file": "speech_9.wav",
|
| 157 |
+
"music": 5.3731204388896e-05,
|
| 158 |
+
"speech": 0.9999462687956111,
|
| 159 |
+
"predicted": "speech",
|
| 160 |
+
"correct": true
|
| 161 |
+
}
|
| 162 |
+
]
|
| 163 |
+
},
|
| 164 |
+
"Speech + Music": {
|
| 165 |
+
"accuracy": 90.0,
|
| 166 |
+
"correct": 9,
|
| 167 |
+
"total": 10,
|
| 168 |
+
"avg_music_score": 0.9352504581212997,
|
| 169 |
+
"avg_speech_score": 0.06474954187870026,
|
| 170 |
+
"files": [
|
| 171 |
+
{
|
| 172 |
+
"file": "speech_and_music_1.wav",
|
| 173 |
+
"music": 0.9999827146530151,
|
| 174 |
+
"speech": 1.728534698486328e-05,
|
| 175 |
+
"predicted": "music",
|
| 176 |
+
"correct": true
|
| 177 |
+
},
|
| 178 |
+
{
|
| 179 |
+
"file": "speech_and_music_10.wav",
|
| 180 |
+
"music": 0.9999879598617554,
|
| 181 |
+
"speech": 1.2040138244628906e-05,
|
| 182 |
+
"predicted": "music",
|
| 183 |
+
"correct": true
|
| 184 |
+
},
|
| 185 |
+
{
|
| 186 |
+
"file": "speech_and_music_2.wav",
|
| 187 |
+
"music": 0.9999924898147583,
|
| 188 |
+
"speech": 7.510185241699219e-06,
|
| 189 |
+
"predicted": "music",
|
| 190 |
+
"correct": true
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"file": "speech_and_music_3wav.wav",
|
| 194 |
+
"music": 0.9999021291732788,
|
| 195 |
+
"speech": 9.78708267211914e-05,
|
| 196 |
+
"predicted": "music",
|
| 197 |
+
"correct": true
|
| 198 |
+
},
|
| 199 |
+
{
|
| 200 |
+
"file": "speech_and_music_4.wav",
|
| 201 |
+
"music": 0.9998724460601807,
|
| 202 |
+
"speech": 0.00012755393981933594,
|
| 203 |
+
"predicted": "music",
|
| 204 |
+
"correct": true
|
| 205 |
+
},
|
| 206 |
+
{
|
| 207 |
+
"file": "speech_and_music_5.wav",
|
| 208 |
+
"music": 0.9999758005142212,
|
| 209 |
+
"speech": 2.4199485778808594e-05,
|
| 210 |
+
"predicted": "music",
|
| 211 |
+
"correct": true
|
| 212 |
+
},
|
| 213 |
+
{
|
| 214 |
+
"file": "speech_and_music_6.wav",
|
| 215 |
+
"music": 0.9999932050704956,
|
| 216 |
+
"speech": 6.794929504394531e-06,
|
| 217 |
+
"predicted": "music",
|
| 218 |
+
"correct": true
|
| 219 |
+
},
|
| 220 |
+
{
|
| 221 |
+
"file": "speech_and_music_7.wav",
|
| 222 |
+
"music": 0.35289058089256287,
|
| 223 |
+
"speech": 0.6471094191074371,
|
| 224 |
+
"predicted": "speech",
|
| 225 |
+
"correct": false
|
| 226 |
+
},
|
| 227 |
+
{
|
| 228 |
+
"file": "speech_and_music_8.wav",
|
| 229 |
+
"music": 0.9999957084655762,
|
| 230 |
+
"speech": 4.291534423828125e-06,
|
| 231 |
+
"predicted": "music",
|
| 232 |
+
"correct": true
|
| 233 |
+
},
|
| 234 |
+
{
|
| 235 |
+
"file": "speech_and_music_9.wav",
|
| 236 |
+
"music": 0.9999115467071533,
|
| 237 |
+
"speech": 8.845329284667969e-05,
|
| 238 |
+
"predicted": "music",
|
| 239 |
+
"correct": true
|
| 240 |
+
}
|
| 241 |
+
]
|
| 242 |
+
}
|
| 243 |
+
}
|
| 244 |
+
}
|