AbstractPhil commited on
Commit
40e5258
·
verified ·
1 Parent(s): 5f49dc6

Upload best.meta.json

Browse files
geom_multitask_mnist_fashion_cifar100_best.meta.json CHANGED
@@ -1,9 +1,9 @@
1
  {
2
- "epoch": 30,
3
- "global_step": 21150,
4
- "avg_loss": 20307.280809507978,
5
- "best_metric": 0.3731333333333334,
6
- "timestamp": "2025-08-25T23:02:35.099929",
7
  "config": {
8
  "seed": 42,
9
  "device": "cuda",
@@ -170,7 +170,7 @@
170
  },
171
  "classification": {
172
  "use_cls": true,
173
- "lambda_cls": 5452.367521599999,
174
  "tau_cls": 0.35
175
  },
176
  "losses": {
@@ -178,24 +178,24 @@
178
  "lambda_rose_mag": 0.05,
179
  "rose_margin": 1.0,
180
  "use_dual_contrastive": true,
181
- "lambda_dual": 1635.5602564799997,
182
  "temp_dual": 0.5,
183
  "use_rose_contrastive": true,
184
- "lambda_rose_contrastive": 1635.5602564799997,
185
  "temp_rose_contrastive": 0.5
186
  },
187
  "train": {
188
  "epochs": 10,
189
  "eval_every_epochs": 1,
190
- "lr": 0.0005,
191
  "grad_clip_norm": 1.0,
192
  "use_anneal": true,
193
  "anneal": {
194
- "tau_start": 0.55,
195
- "drift_start": 0.05,
196
- "lambda_cls_start": 1.0,
197
- "lambda_dual_start": 0.15,
198
- "lambda_rose_contrastive_start": 0.15
199
  }
200
  },
201
  "io": {
@@ -228,16 +228,16 @@
228
  },
229
  "per_task": {
230
  "mnist": {
231
- "top1": 0.6417,
232
- "top3": 0.9486
233
  },
234
  "fashion": {
235
- "top1": 0.4602,
236
- "top3": 0.8244
237
  },
238
  "cifar100": {
239
- "top1": 0.0175,
240
- "top3": 0.0411
241
  }
242
  }
243
  }
 
1
  {
2
+ "epoch": 39,
3
+ "global_step": 27495,
4
+ "avg_loss": 192415.9321143617,
5
+ "best_metric": 0.38010000000000005,
6
+ "timestamp": "2025-08-25T23:09:02.046735",
7
  "config": {
8
  "seed": 42,
9
  "device": "cuda",
 
170
  },
171
  "classification": {
172
  "use_cls": true,
173
+ "lambda_cls": 60751.522849126406,
174
  "tau_cls": 0.35
175
  },
176
  "losses": {
 
178
  "lambda_rose_mag": 0.05,
179
  "rose_margin": 1.0,
180
  "use_dual_contrastive": true,
181
+ "lambda_dual": 3845.2381550079995,
182
  "temp_dual": 0.5,
183
  "use_rose_contrastive": true,
184
+ "lambda_rose_contrastive": 653.939486351361,
185
  "temp_rose_contrastive": 0.5
186
  },
187
  "train": {
188
  "epochs": 10,
189
  "eval_every_epochs": 1,
190
+ "lr": 0.001,
191
  "grad_clip_norm": 1.0,
192
  "use_anneal": true,
193
  "anneal": {
194
+ "tau_start": 0.25,
195
+ "drift_start": 0.01,
196
+ "lambda_cls_start": 0.71,
197
+ "lambda_dual_start": 0.25,
198
+ "lambda_rose_contrastive_start": 0.2915
199
  }
200
  },
201
  "io": {
 
228
  },
229
  "per_task": {
230
  "mnist": {
231
+ "top1": 0.7088,
232
+ "top3": 0.9593
233
  },
234
  "fashion": {
235
+ "top1": 0.4162,
236
+ "top3": 0.8973
237
  },
238
  "cifar100": {
239
+ "top1": 0.0153,
240
+ "top3": 0.0458
241
  }
242
  }
243
  }