Hugging Face's logo

ModelsLab
/
midashenglm-gen-wer-lora

Text-to-Audio
PEFT
Safetensors
audio-generation
lora
flow-matching
Model card Files
xet
 Community
adhikjoshi commited on
Commit
dc6bbfc
·
verified ·
1 Parent(s): 30ff3c2

Upload history.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. history.json +101 -13
history.json CHANGED
@@ -196,29 +196,117 @@
196
  },
197
  "accepted": true,
198
  "stale": 1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
199
  }
200
  ],
201
  "best": {
202
- "score": 0.15797730777412652,
203
- "step": 750,
204
  "losses": {
205
- "speech": 0.09360956773161888,
206
- "sfx": 0.20055881701409817,
207
- "music": 0.16840521432459354,
208
- "ambience": 0.20233918726444244,
209
- "mixed": 0.12497375253587961
210
  },
211
  "deltas": {
212
- "speech": -0.021005277917062894,
213
- "sfx": -0.014920306525608239,
214
- "music": -0.012956909201448576,
215
- "ambience": -0.00606670653317617,
216
- "mixed": -0.006680225822419227
217
  },
218
  "stop_ratio": {
219
  "speech": 0.9989583333333334,
220
  "sfx": 0.9924806749622928,
221
- "music": 0.9987702546296297,
222
  "ambience": 1.0,
223
  "mixed": 0.9989255536130536
224
  }
 
196
  },
197
  "accepted": true,
198
  "stale": 1
199
+ },
200
+ {
201
+ "step": 1000,
202
+ "losses": {
203
+ "speech": 0.09348684921860695,
204
+ "sfx": 0.2005684357136488,
205
+ "music": 0.16837280616164207,
206
+ "ambience": 0.202393414452672,
207
+ "mixed": 0.1249053431674838
208
+ },
209
+ "deltas": {
210
+ "speech": -0.022288702031198267,
211
+ "sfx": -0.014873062601127524,
212
+ "music": -0.013146857318286962,
213
+ "ambience": -0.0058003310054242875,
214
+ "mixed": -0.007223959022292884
215
+ },
216
+ "mean": 0.15794536974281073,
217
+ "stop_ratio": {
218
+ "speech": 0.9989583333333334,
219
+ "sfx": 0.9924806749622928,
220
+ "music": 0.9987702546296297,
221
+ "ambience": 1.0,
222
+ "mixed": 0.9989255536130536
223
+ },
224
+ "stop_shrink": {
225
+ "speech": 0.0,
226
+ "sfx": 0.0,
227
+ "music": 0.0,
228
+ "ambience": 0.0,
229
+ "mixed": 0.0
230
+ },
231
+ "gen_lengths": {
232
+ "sfx-gravel": 10.0,
233
+ "sfx-fire": 10.0,
234
+ "ambience-rain": 10.0
235
+ },
236
+ "gen_shrink": {
237
+ "sfx-gravel": 0.0,
238
+ "sfx-fire": 0.0,
239
+ "ambience-rain": 0.0
240
+ },
241
+ "accepted": true,
242
+ "stale": 2
243
+ },
244
+ {
245
+ "step": 1250,
246
+ "losses": {
247
+ "speech": 0.09341771434992552,
248
+ "sfx": 0.20048575475811958,
249
+ "music": 0.16836701519787312,
250
+ "ambience": 0.20210730656981468,
251
+ "mixed": 0.12483794149011374
252
+ },
253
+ "deltas": {
254
+ "speech": -0.023011733588667538,
255
+ "sfx": -0.015279164569325519,
256
+ "music": -0.013180798849136468,
257
+ "ambience": -0.007205754018829721,
258
+ "mixed": -0.0077596828649870355
259
+ },
260
+ "mean": 0.15784314647316933,
261
+ "stop_ratio": {
262
+ "speech": 0.9989583333333334,
263
+ "sfx": 0.9924806749622928,
264
+ "music": 0.9984447337962963,
265
+ "ambience": 1.0,
266
+ "mixed": 0.9989255536130536
267
+ },
268
+ "stop_shrink": {
269
+ "speech": 0.0,
270
+ "sfx": 0.0,
271
+ "music": 0.0003259216339538286,
272
+ "ambience": 0.0,
273
+ "mixed": 0.0
274
+ },
275
+ "gen_lengths": {
276
+ "sfx-gravel": 10.0,
277
+ "sfx-fire": 10.0,
278
+ "ambience-rain": 9.8
279
+ },
280
+ "gen_shrink": {
281
+ "sfx-gravel": 0.0,
282
+ "sfx-fire": 0.0,
283
+ "ambience-rain": 0.019999999999999928
284
+ },
285
+ "accepted": true,
286
+ "stale": 3
287
  }
288
  ],
289
  "best": {
290
+ "score": 0.15784314647316933,
291
+ "step": 1250,
292
  "losses": {
293
+ "speech": 0.09341771434992552,
294
+ "sfx": 0.20048575475811958,
295
+ "music": 0.16836701519787312,
296
+ "ambience": 0.20210730656981468,
297
+ "mixed": 0.12483794149011374
298
  },
299
  "deltas": {
300
+ "speech": -0.023011733588667538,
301
+ "sfx": -0.015279164569325519,
302
+ "music": -0.013180798849136468,
303
+ "ambience": -0.007205754018829721,
304
+ "mixed": -0.0077596828649870355
305
  },
306
  "stop_ratio": {
307
  "speech": 0.9989583333333334,
308
  "sfx": 0.9924806749622928,
309
+ "music": 0.9984447337962963,
310
  "ambience": 1.0,
311
  "mixed": 0.9989255536130536
312
  }