Leensman commited on
Commit
be24bcb
1 Parent(s): 444cfe5

Upload 7 files

Browse files

Serseg-B0 with DConv decoder for easyportrait dataset

The model repo is here https://github.com/HolyBayes/FastSeg/blob/main/README.md

Files changed (7) hide show
  1. config.json +82 -0
  2. model.safetensors +3 -0
  3. optimizer.pt +3 -0
  4. rng_state.pth +3 -0
  5. scheduler.pt +3 -0
  6. trainer_state.json +1541 -0
  7. training_args.bin +3 -0
config.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "../checkpoints/serseg_w_decoder/",
3
+ "abg": false,
4
+ "add_depth_channel": false,
5
+ "afn": false,
6
+ "architectures": [
7
+ "SersegformerForSemanticSegmentation"
8
+ ],
9
+ "attention_probs_dropout_prob": 0.0,
10
+ "classifier_dropout_prob": 0.1,
11
+ "critic": false,
12
+ "dam": false,
13
+ "dbn": false,
14
+ "decoder_hidden_size": 256,
15
+ "depths": [
16
+ 2,
17
+ 2,
18
+ 2,
19
+ 2
20
+ ],
21
+ "downsampling_rates": [
22
+ 1,
23
+ 4,
24
+ 8,
25
+ 16
26
+ ],
27
+ "drop_path_rate": 0.1,
28
+ "hidden_act": "gelu",
29
+ "hidden_dropout_prob": 0.0,
30
+ "hidden_sizes": [
31
+ 32,
32
+ 64,
33
+ 160,
34
+ 256
35
+ ],
36
+ "id2label": {
37
+ "0": "LABEL_0"
38
+ },
39
+ "image_size": 512,
40
+ "initializer_range": 0.02,
41
+ "label2id": {
42
+ "LABEL_0": 0
43
+ },
44
+ "layer_norm_eps": 1e-06,
45
+ "mlp_ratios": [
46
+ 4,
47
+ 4,
48
+ 4,
49
+ 4
50
+ ],
51
+ "model_type": "sersegformer",
52
+ "num_attention_heads": [
53
+ 1,
54
+ 2,
55
+ 5,
56
+ 8
57
+ ],
58
+ "num_channels": 3,
59
+ "num_encoder_blocks": 4,
60
+ "patch_sizes": [
61
+ 7,
62
+ 3,
63
+ 3,
64
+ 3
65
+ ],
66
+ "reshape_last_stage": true,
67
+ "sr_ratios": [
68
+ 8,
69
+ 4,
70
+ 2,
71
+ 1
72
+ ],
73
+ "strides": [
74
+ 4,
75
+ 2,
76
+ 2,
77
+ 2
78
+ ],
79
+ "torch_dtype": "float32",
80
+ "transformers_version": "4.37.1",
81
+ "upsample": true
82
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8574cb86343640341bd8327435c088653b37c2fd94cd02b8a3e79f0d17fb41e8
3
+ size 15624300
optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:103e71fa9539fe954fb924fc8711bc3ac443474890b858138227d25189426079
3
+ size 31315002
rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:66dffff880309c3ffd4353adfa35078070a3dd0eceaeed5843012743b07f5485
3
+ size 14180
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:07d269727d16e814cf8bc7a7a09468a8d9b1d31aa4b9c95bf8e4dbc1c031e0b2
3
+ size 1000
trainer_state.json ADDED
@@ -0,0 +1,1541 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.9827606730601367,
3
+ "best_model_checkpoint": "../results/SersegformerForSemanticSegmentation_decoder/checkpoint-21500",
4
+ "epoch": 15.357142857142858,
5
+ "eval_steps": 500,
6
+ "global_step": 21500,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.07,
13
+ "learning_rate": 4e-06,
14
+ "loss": 0.0169,
15
+ "step": 100
16
+ },
17
+ {
18
+ "epoch": 0.14,
19
+ "learning_rate": 8e-06,
20
+ "loss": 0.0171,
21
+ "step": 200
22
+ },
23
+ {
24
+ "epoch": 0.21,
25
+ "learning_rate": 1.2e-05,
26
+ "loss": 0.0182,
27
+ "step": 300
28
+ },
29
+ {
30
+ "epoch": 0.29,
31
+ "learning_rate": 1.6e-05,
32
+ "loss": 0.0173,
33
+ "step": 400
34
+ },
35
+ {
36
+ "epoch": 0.36,
37
+ "learning_rate": 1.9999999999999998e-05,
38
+ "loss": 0.017,
39
+ "step": 500
40
+ },
41
+ {
42
+ "epoch": 0.36,
43
+ "eval_iou": 0.9819581976123885,
44
+ "step": 500
45
+ },
46
+ {
47
+ "epoch": 0.43,
48
+ "learning_rate": 2.4e-05,
49
+ "loss": 0.0169,
50
+ "step": 600
51
+ },
52
+ {
53
+ "epoch": 0.5,
54
+ "learning_rate": 2.8e-05,
55
+ "loss": 0.0163,
56
+ "step": 700
57
+ },
58
+ {
59
+ "epoch": 0.57,
60
+ "learning_rate": 3.2e-05,
61
+ "loss": 0.018,
62
+ "step": 800
63
+ },
64
+ {
65
+ "epoch": 0.64,
66
+ "learning_rate": 3.6e-05,
67
+ "loss": 0.0175,
68
+ "step": 900
69
+ },
70
+ {
71
+ "epoch": 0.71,
72
+ "learning_rate": 3.9999999999999996e-05,
73
+ "loss": 0.0167,
74
+ "step": 1000
75
+ },
76
+ {
77
+ "epoch": 0.71,
78
+ "eval_iou": 0.9816833290399289,
79
+ "step": 1000
80
+ },
81
+ {
82
+ "epoch": 0.79,
83
+ "learning_rate": 4.4e-05,
84
+ "loss": 0.0174,
85
+ "step": 1100
86
+ },
87
+ {
88
+ "epoch": 0.86,
89
+ "learning_rate": 4.8e-05,
90
+ "loss": 0.0171,
91
+ "step": 1200
92
+ },
93
+ {
94
+ "epoch": 0.93,
95
+ "learning_rate": 5.2000000000000004e-05,
96
+ "loss": 0.0169,
97
+ "step": 1300
98
+ },
99
+ {
100
+ "epoch": 1.0,
101
+ "learning_rate": 5.6e-05,
102
+ "loss": 0.0171,
103
+ "step": 1400
104
+ },
105
+ {
106
+ "epoch": 1.07,
107
+ "learning_rate": 6e-05,
108
+ "loss": 0.0175,
109
+ "step": 1500
110
+ },
111
+ {
112
+ "epoch": 1.07,
113
+ "eval_iou": 0.9816984183648053,
114
+ "step": 1500
115
+ },
116
+ {
117
+ "epoch": 1.14,
118
+ "learning_rate": 5.98520987654321e-05,
119
+ "loss": 0.0178,
120
+ "step": 1600
121
+ },
122
+ {
123
+ "epoch": 1.21,
124
+ "learning_rate": 5.970419753086419e-05,
125
+ "loss": 0.0184,
126
+ "step": 1700
127
+ },
128
+ {
129
+ "epoch": 1.29,
130
+ "learning_rate": 5.95562962962963e-05,
131
+ "loss": 0.018,
132
+ "step": 1800
133
+ },
134
+ {
135
+ "epoch": 1.36,
136
+ "learning_rate": 5.94083950617284e-05,
137
+ "loss": 0.0186,
138
+ "step": 1900
139
+ },
140
+ {
141
+ "epoch": 1.43,
142
+ "learning_rate": 5.926049382716049e-05,
143
+ "loss": 0.0205,
144
+ "step": 2000
145
+ },
146
+ {
147
+ "epoch": 1.43,
148
+ "eval_iou": 0.9816799380031287,
149
+ "step": 2000
150
+ },
151
+ {
152
+ "epoch": 1.5,
153
+ "learning_rate": 5.911259259259259e-05,
154
+ "loss": 0.0194,
155
+ "step": 2100
156
+ },
157
+ {
158
+ "epoch": 1.57,
159
+ "learning_rate": 5.8964691358024696e-05,
160
+ "loss": 0.018,
161
+ "step": 2200
162
+ },
163
+ {
164
+ "epoch": 1.64,
165
+ "learning_rate": 5.8816790123456795e-05,
166
+ "loss": 0.0192,
167
+ "step": 2300
168
+ },
169
+ {
170
+ "epoch": 1.71,
171
+ "learning_rate": 5.866888888888889e-05,
172
+ "loss": 0.0186,
173
+ "step": 2400
174
+ },
175
+ {
176
+ "epoch": 1.79,
177
+ "learning_rate": 5.852098765432099e-05,
178
+ "loss": 0.0179,
179
+ "step": 2500
180
+ },
181
+ {
182
+ "epoch": 1.79,
183
+ "eval_iou": 0.9817095761205635,
184
+ "step": 2500
185
+ },
186
+ {
187
+ "epoch": 1.86,
188
+ "learning_rate": 5.837308641975309e-05,
189
+ "loss": 0.0195,
190
+ "step": 2600
191
+ },
192
+ {
193
+ "epoch": 1.93,
194
+ "learning_rate": 5.8225185185185185e-05,
195
+ "loss": 0.0189,
196
+ "step": 2700
197
+ },
198
+ {
199
+ "epoch": 2.0,
200
+ "learning_rate": 5.8077283950617285e-05,
201
+ "loss": 0.0187,
202
+ "step": 2800
203
+ },
204
+ {
205
+ "epoch": 2.07,
206
+ "learning_rate": 5.7929382716049384e-05,
207
+ "loss": 0.0179,
208
+ "step": 2900
209
+ },
210
+ {
211
+ "epoch": 2.14,
212
+ "learning_rate": 5.778148148148149e-05,
213
+ "loss": 0.018,
214
+ "step": 3000
215
+ },
216
+ {
217
+ "epoch": 2.14,
218
+ "eval_iou": 0.9814854831087823,
219
+ "step": 3000
220
+ },
221
+ {
222
+ "epoch": 2.21,
223
+ "learning_rate": 5.763358024691358e-05,
224
+ "loss": 0.0192,
225
+ "step": 3100
226
+ },
227
+ {
228
+ "epoch": 2.29,
229
+ "learning_rate": 5.748567901234568e-05,
230
+ "loss": 0.0168,
231
+ "step": 3200
232
+ },
233
+ {
234
+ "epoch": 2.36,
235
+ "learning_rate": 5.733777777777778e-05,
236
+ "loss": 0.0179,
237
+ "step": 3300
238
+ },
239
+ {
240
+ "epoch": 2.43,
241
+ "learning_rate": 5.718987654320987e-05,
242
+ "loss": 0.0178,
243
+ "step": 3400
244
+ },
245
+ {
246
+ "epoch": 2.5,
247
+ "learning_rate": 5.704197530864198e-05,
248
+ "loss": 0.0182,
249
+ "step": 3500
250
+ },
251
+ {
252
+ "epoch": 2.5,
253
+ "eval_iou": 0.9810651301168928,
254
+ "step": 3500
255
+ },
256
+ {
257
+ "epoch": 2.57,
258
+ "learning_rate": 5.689407407407408e-05,
259
+ "loss": 0.0181,
260
+ "step": 3600
261
+ },
262
+ {
263
+ "epoch": 2.64,
264
+ "learning_rate": 5.674617283950617e-05,
265
+ "loss": 0.019,
266
+ "step": 3700
267
+ },
268
+ {
269
+ "epoch": 2.71,
270
+ "learning_rate": 5.659827160493827e-05,
271
+ "loss": 0.0181,
272
+ "step": 3800
273
+ },
274
+ {
275
+ "epoch": 2.79,
276
+ "learning_rate": 5.6450370370370376e-05,
277
+ "loss": 0.0179,
278
+ "step": 3900
279
+ },
280
+ {
281
+ "epoch": 2.86,
282
+ "learning_rate": 5.630246913580247e-05,
283
+ "loss": 0.0181,
284
+ "step": 4000
285
+ },
286
+ {
287
+ "epoch": 2.86,
288
+ "eval_iou": 0.9813077029059915,
289
+ "step": 4000
290
+ },
291
+ {
292
+ "epoch": 2.93,
293
+ "learning_rate": 5.615456790123457e-05,
294
+ "loss": 0.0176,
295
+ "step": 4100
296
+ },
297
+ {
298
+ "epoch": 3.0,
299
+ "learning_rate": 5.600666666666667e-05,
300
+ "loss": 0.0178,
301
+ "step": 4200
302
+ },
303
+ {
304
+ "epoch": 3.07,
305
+ "learning_rate": 5.585876543209877e-05,
306
+ "loss": 0.0185,
307
+ "step": 4300
308
+ },
309
+ {
310
+ "epoch": 3.14,
311
+ "learning_rate": 5.5710864197530865e-05,
312
+ "loss": 0.0172,
313
+ "step": 4400
314
+ },
315
+ {
316
+ "epoch": 3.21,
317
+ "learning_rate": 5.5562962962962965e-05,
318
+ "loss": 0.0169,
319
+ "step": 4500
320
+ },
321
+ {
322
+ "epoch": 3.21,
323
+ "eval_iou": 0.981582362277835,
324
+ "step": 4500
325
+ },
326
+ {
327
+ "epoch": 3.29,
328
+ "learning_rate": 5.5415061728395064e-05,
329
+ "loss": 0.0174,
330
+ "step": 4600
331
+ },
332
+ {
333
+ "epoch": 3.36,
334
+ "learning_rate": 5.526716049382716e-05,
335
+ "loss": 0.0188,
336
+ "step": 4700
337
+ },
338
+ {
339
+ "epoch": 3.43,
340
+ "learning_rate": 5.511925925925926e-05,
341
+ "loss": 0.0175,
342
+ "step": 4800
343
+ },
344
+ {
345
+ "epoch": 3.5,
346
+ "learning_rate": 5.497135802469136e-05,
347
+ "loss": 0.0177,
348
+ "step": 4900
349
+ },
350
+ {
351
+ "epoch": 3.57,
352
+ "learning_rate": 5.482345679012346e-05,
353
+ "loss": 0.0179,
354
+ "step": 5000
355
+ },
356
+ {
357
+ "epoch": 3.57,
358
+ "eval_iou": 0.981942740725536,
359
+ "step": 5000
360
+ },
361
+ {
362
+ "epoch": 3.64,
363
+ "learning_rate": 5.467555555555555e-05,
364
+ "loss": 0.0195,
365
+ "step": 5100
366
+ },
367
+ {
368
+ "epoch": 3.71,
369
+ "learning_rate": 5.452765432098766e-05,
370
+ "loss": 0.0202,
371
+ "step": 5200
372
+ },
373
+ {
374
+ "epoch": 3.79,
375
+ "learning_rate": 5.437975308641975e-05,
376
+ "loss": 0.0198,
377
+ "step": 5300
378
+ },
379
+ {
380
+ "epoch": 3.86,
381
+ "learning_rate": 5.423185185185185e-05,
382
+ "loss": 0.0198,
383
+ "step": 5400
384
+ },
385
+ {
386
+ "epoch": 3.93,
387
+ "learning_rate": 5.408395061728395e-05,
388
+ "loss": 0.0177,
389
+ "step": 5500
390
+ },
391
+ {
392
+ "epoch": 3.93,
393
+ "eval_iou": 0.9819124785124087,
394
+ "step": 5500
395
+ },
396
+ {
397
+ "epoch": 4.0,
398
+ "learning_rate": 5.3936049382716056e-05,
399
+ "loss": 0.0197,
400
+ "step": 5600
401
+ },
402
+ {
403
+ "epoch": 4.07,
404
+ "learning_rate": 5.378814814814815e-05,
405
+ "loss": 0.0186,
406
+ "step": 5700
407
+ },
408
+ {
409
+ "epoch": 4.14,
410
+ "learning_rate": 5.364024691358025e-05,
411
+ "loss": 0.018,
412
+ "step": 5800
413
+ },
414
+ {
415
+ "epoch": 4.21,
416
+ "learning_rate": 5.349234567901235e-05,
417
+ "loss": 0.0188,
418
+ "step": 5900
419
+ },
420
+ {
421
+ "epoch": 4.29,
422
+ "learning_rate": 5.3344444444444446e-05,
423
+ "loss": 0.0187,
424
+ "step": 6000
425
+ },
426
+ {
427
+ "epoch": 4.29,
428
+ "eval_iou": 0.9817866309016359,
429
+ "step": 6000
430
+ },
431
+ {
432
+ "epoch": 4.36,
433
+ "learning_rate": 5.3196543209876545e-05,
434
+ "loss": 0.019,
435
+ "step": 6100
436
+ },
437
+ {
438
+ "epoch": 4.43,
439
+ "learning_rate": 5.3048641975308645e-05,
440
+ "loss": 0.018,
441
+ "step": 6200
442
+ },
443
+ {
444
+ "epoch": 4.5,
445
+ "learning_rate": 5.2900740740740744e-05,
446
+ "loss": 0.0193,
447
+ "step": 6300
448
+ },
449
+ {
450
+ "epoch": 4.57,
451
+ "learning_rate": 5.2752839506172836e-05,
452
+ "loss": 0.0187,
453
+ "step": 6400
454
+ },
455
+ {
456
+ "epoch": 4.64,
457
+ "learning_rate": 5.260493827160494e-05,
458
+ "loss": 0.0182,
459
+ "step": 6500
460
+ },
461
+ {
462
+ "epoch": 4.64,
463
+ "eval_iou": 0.9818342994241154,
464
+ "step": 6500
465
+ },
466
+ {
467
+ "epoch": 4.71,
468
+ "learning_rate": 5.245703703703704e-05,
469
+ "loss": 0.0184,
470
+ "step": 6600
471
+ },
472
+ {
473
+ "epoch": 4.79,
474
+ "learning_rate": 5.2309135802469134e-05,
475
+ "loss": 0.0182,
476
+ "step": 6700
477
+ },
478
+ {
479
+ "epoch": 4.86,
480
+ "learning_rate": 5.216123456790123e-05,
481
+ "loss": 0.0175,
482
+ "step": 6800
483
+ },
484
+ {
485
+ "epoch": 4.93,
486
+ "learning_rate": 5.201333333333334e-05,
487
+ "loss": 0.0178,
488
+ "step": 6900
489
+ },
490
+ {
491
+ "epoch": 5.0,
492
+ "learning_rate": 5.186543209876544e-05,
493
+ "loss": 0.0178,
494
+ "step": 7000
495
+ },
496
+ {
497
+ "epoch": 5.0,
498
+ "eval_iou": 0.9820612844298867,
499
+ "step": 7000
500
+ },
501
+ {
502
+ "epoch": 5.0,
503
+ "eval_iou": 0.9820612844298867,
504
+ "step": 7000
505
+ },
506
+ {
507
+ "epoch": 5.07,
508
+ "learning_rate": 5.171753086419753e-05,
509
+ "loss": 0.018,
510
+ "step": 7100
511
+ },
512
+ {
513
+ "epoch": 5.14,
514
+ "learning_rate": 5.156962962962963e-05,
515
+ "loss": 0.0196,
516
+ "step": 7200
517
+ },
518
+ {
519
+ "epoch": 5.21,
520
+ "learning_rate": 5.142172839506173e-05,
521
+ "loss": 0.0198,
522
+ "step": 7300
523
+ },
524
+ {
525
+ "epoch": 5.29,
526
+ "learning_rate": 5.127382716049383e-05,
527
+ "loss": 0.018,
528
+ "step": 7400
529
+ },
530
+ {
531
+ "epoch": 5.36,
532
+ "learning_rate": 5.112592592592593e-05,
533
+ "loss": 0.0191,
534
+ "step": 7500
535
+ },
536
+ {
537
+ "epoch": 5.36,
538
+ "eval_iou": 0.9817404121744866,
539
+ "step": 7500
540
+ },
541
+ {
542
+ "epoch": 5.43,
543
+ "learning_rate": 5.097802469135803e-05,
544
+ "loss": 0.0174,
545
+ "step": 7600
546
+ },
547
+ {
548
+ "epoch": 5.5,
549
+ "learning_rate": 5.0830123456790126e-05,
550
+ "loss": 0.0174,
551
+ "step": 7700
552
+ },
553
+ {
554
+ "epoch": 5.57,
555
+ "learning_rate": 5.0682222222222225e-05,
556
+ "loss": 0.019,
557
+ "step": 7800
558
+ },
559
+ {
560
+ "epoch": 5.64,
561
+ "learning_rate": 5.0534320987654324e-05,
562
+ "loss": 0.0186,
563
+ "step": 7900
564
+ },
565
+ {
566
+ "epoch": 5.71,
567
+ "learning_rate": 5.038641975308642e-05,
568
+ "loss": 0.018,
569
+ "step": 8000
570
+ },
571
+ {
572
+ "epoch": 5.71,
573
+ "eval_iou": 0.981840242357815,
574
+ "step": 8000
575
+ },
576
+ {
577
+ "epoch": 5.79,
578
+ "learning_rate": 5.0238518518518516e-05,
579
+ "loss": 0.0174,
580
+ "step": 8100
581
+ },
582
+ {
583
+ "epoch": 5.86,
584
+ "learning_rate": 5.009061728395062e-05,
585
+ "loss": 0.0179,
586
+ "step": 8200
587
+ },
588
+ {
589
+ "epoch": 5.93,
590
+ "learning_rate": 4.994271604938272e-05,
591
+ "loss": 0.0182,
592
+ "step": 8300
593
+ },
594
+ {
595
+ "epoch": 6.0,
596
+ "learning_rate": 4.9794814814814814e-05,
597
+ "loss": 0.0181,
598
+ "step": 8400
599
+ },
600
+ {
601
+ "epoch": 6.07,
602
+ "learning_rate": 4.964691358024691e-05,
603
+ "loss": 0.0167,
604
+ "step": 8500
605
+ },
606
+ {
607
+ "epoch": 6.07,
608
+ "eval_iou": 0.9818143762794196,
609
+ "step": 8500
610
+ },
611
+ {
612
+ "epoch": 6.14,
613
+ "learning_rate": 4.949901234567901e-05,
614
+ "loss": 0.0173,
615
+ "step": 8600
616
+ },
617
+ {
618
+ "epoch": 6.21,
619
+ "learning_rate": 4.935111111111111e-05,
620
+ "loss": 0.0184,
621
+ "step": 8700
622
+ },
623
+ {
624
+ "epoch": 6.29,
625
+ "learning_rate": 4.920320987654321e-05,
626
+ "loss": 0.0188,
627
+ "step": 8800
628
+ },
629
+ {
630
+ "epoch": 6.36,
631
+ "learning_rate": 4.905530864197531e-05,
632
+ "loss": 0.0183,
633
+ "step": 8900
634
+ },
635
+ {
636
+ "epoch": 6.43,
637
+ "learning_rate": 4.8907407407407416e-05,
638
+ "loss": 0.0176,
639
+ "step": 9000
640
+ },
641
+ {
642
+ "epoch": 6.43,
643
+ "eval_iou": 0.9819974823325288,
644
+ "step": 9000
645
+ },
646
+ {
647
+ "epoch": 6.5,
648
+ "learning_rate": 4.875950617283951e-05,
649
+ "loss": 0.0195,
650
+ "step": 9100
651
+ },
652
+ {
653
+ "epoch": 6.57,
654
+ "learning_rate": 4.861160493827161e-05,
655
+ "loss": 0.0176,
656
+ "step": 9200
657
+ },
658
+ {
659
+ "epoch": 6.64,
660
+ "learning_rate": 4.84637037037037e-05,
661
+ "loss": 0.0186,
662
+ "step": 9300
663
+ },
664
+ {
665
+ "epoch": 6.71,
666
+ "learning_rate": 4.8315802469135806e-05,
667
+ "loss": 0.0194,
668
+ "step": 9400
669
+ },
670
+ {
671
+ "epoch": 6.79,
672
+ "learning_rate": 4.8167901234567905e-05,
673
+ "loss": 0.0179,
674
+ "step": 9500
675
+ },
676
+ {
677
+ "epoch": 6.79,
678
+ "eval_iou": 0.9821633836802315,
679
+ "step": 9500
680
+ },
681
+ {
682
+ "epoch": 6.86,
683
+ "learning_rate": 4.8020000000000004e-05,
684
+ "loss": 0.0167,
685
+ "step": 9600
686
+ },
687
+ {
688
+ "epoch": 6.93,
689
+ "learning_rate": 4.78720987654321e-05,
690
+ "loss": 0.0175,
691
+ "step": 9700
692
+ },
693
+ {
694
+ "epoch": 7.0,
695
+ "learning_rate": 4.7724197530864196e-05,
696
+ "loss": 0.0171,
697
+ "step": 9800
698
+ },
699
+ {
700
+ "epoch": 7.07,
701
+ "learning_rate": 4.75762962962963e-05,
702
+ "loss": 0.0191,
703
+ "step": 9900
704
+ },
705
+ {
706
+ "epoch": 7.14,
707
+ "learning_rate": 4.7428395061728395e-05,
708
+ "loss": 0.0188,
709
+ "step": 10000
710
+ },
711
+ {
712
+ "epoch": 7.14,
713
+ "eval_iou": 0.9819790539788265,
714
+ "step": 10000
715
+ },
716
+ {
717
+ "epoch": 7.21,
718
+ "learning_rate": 4.7280493827160494e-05,
719
+ "loss": 0.0179,
720
+ "step": 10100
721
+ },
722
+ {
723
+ "epoch": 7.29,
724
+ "learning_rate": 4.713259259259259e-05,
725
+ "loss": 0.0171,
726
+ "step": 10200
727
+ },
728
+ {
729
+ "epoch": 7.36,
730
+ "learning_rate": 4.69846913580247e-05,
731
+ "loss": 0.0178,
732
+ "step": 10300
733
+ },
734
+ {
735
+ "epoch": 7.43,
736
+ "learning_rate": 4.683679012345679e-05,
737
+ "loss": 0.0179,
738
+ "step": 10400
739
+ },
740
+ {
741
+ "epoch": 7.5,
742
+ "learning_rate": 4.668888888888889e-05,
743
+ "loss": 0.0183,
744
+ "step": 10500
745
+ },
746
+ {
747
+ "epoch": 7.5,
748
+ "eval_iou": 0.9820494610889285,
749
+ "step": 10500
750
+ },
751
+ {
752
+ "epoch": 7.57,
753
+ "learning_rate": 4.654098765432098e-05,
754
+ "loss": 0.0184,
755
+ "step": 10600
756
+ },
757
+ {
758
+ "epoch": 7.64,
759
+ "learning_rate": 4.639308641975309e-05,
760
+ "loss": 0.0186,
761
+ "step": 10700
762
+ },
763
+ {
764
+ "epoch": 7.71,
765
+ "learning_rate": 4.624518518518519e-05,
766
+ "loss": 0.0178,
767
+ "step": 10800
768
+ },
769
+ {
770
+ "epoch": 7.79,
771
+ "learning_rate": 4.609728395061729e-05,
772
+ "loss": 0.0176,
773
+ "step": 10900
774
+ },
775
+ {
776
+ "epoch": 7.86,
777
+ "learning_rate": 4.594938271604939e-05,
778
+ "loss": 0.0175,
779
+ "step": 11000
780
+ },
781
+ {
782
+ "epoch": 7.86,
783
+ "eval_iou": 0.9821178334600785,
784
+ "step": 11000
785
+ },
786
+ {
787
+ "epoch": 7.93,
788
+ "learning_rate": 4.5801481481481486e-05,
789
+ "loss": 0.019,
790
+ "step": 11100
791
+ },
792
+ {
793
+ "epoch": 8.0,
794
+ "learning_rate": 4.5653580246913585e-05,
795
+ "loss": 0.0181,
796
+ "step": 11200
797
+ },
798
+ {
799
+ "epoch": 8.07,
800
+ "learning_rate": 4.550567901234568e-05,
801
+ "loss": 0.0176,
802
+ "step": 11300
803
+ },
804
+ {
805
+ "epoch": 8.14,
806
+ "learning_rate": 4.535777777777778e-05,
807
+ "loss": 0.0171,
808
+ "step": 11400
809
+ },
810
+ {
811
+ "epoch": 8.21,
812
+ "learning_rate": 4.5209876543209876e-05,
813
+ "loss": 0.0178,
814
+ "step": 11500
815
+ },
816
+ {
817
+ "epoch": 8.21,
818
+ "eval_iou": 0.9818687801267586,
819
+ "step": 11500
820
+ },
821
+ {
822
+ "epoch": 8.29,
823
+ "learning_rate": 4.506197530864198e-05,
824
+ "loss": 0.0173,
825
+ "step": 11600
826
+ },
827
+ {
828
+ "epoch": 8.36,
829
+ "learning_rate": 4.491407407407408e-05,
830
+ "loss": 0.0179,
831
+ "step": 11700
832
+ },
833
+ {
834
+ "epoch": 8.43,
835
+ "learning_rate": 4.4766172839506174e-05,
836
+ "loss": 0.0182,
837
+ "step": 11800
838
+ },
839
+ {
840
+ "epoch": 8.5,
841
+ "learning_rate": 4.4618271604938266e-05,
842
+ "loss": 0.0185,
843
+ "step": 11900
844
+ },
845
+ {
846
+ "epoch": 8.57,
847
+ "learning_rate": 4.447037037037037e-05,
848
+ "loss": 0.0175,
849
+ "step": 12000
850
+ },
851
+ {
852
+ "epoch": 8.57,
853
+ "eval_iou": 0.9821109941192702,
854
+ "step": 12000
855
+ },
856
+ {
857
+ "epoch": 8.64,
858
+ "learning_rate": 4.432246913580247e-05,
859
+ "loss": 0.0171,
860
+ "step": 12100
861
+ },
862
+ {
863
+ "epoch": 8.71,
864
+ "learning_rate": 4.417456790123457e-05,
865
+ "loss": 0.0178,
866
+ "step": 12200
867
+ },
868
+ {
869
+ "epoch": 8.79,
870
+ "learning_rate": 4.402666666666667e-05,
871
+ "loss": 0.0179,
872
+ "step": 12300
873
+ },
874
+ {
875
+ "epoch": 8.86,
876
+ "learning_rate": 4.387876543209877e-05,
877
+ "loss": 0.0178,
878
+ "step": 12400
879
+ },
880
+ {
881
+ "epoch": 8.93,
882
+ "learning_rate": 4.373086419753087e-05,
883
+ "loss": 0.0177,
884
+ "step": 12500
885
+ },
886
+ {
887
+ "epoch": 8.93,
888
+ "eval_iou": 0.9821157449600744,
889
+ "step": 12500
890
+ },
891
+ {
892
+ "epoch": 9.0,
893
+ "learning_rate": 4.358296296296296e-05,
894
+ "loss": 0.0171,
895
+ "step": 12600
896
+ },
897
+ {
898
+ "epoch": 9.07,
899
+ "learning_rate": 4.343506172839506e-05,
900
+ "loss": 0.0172,
901
+ "step": 12700
902
+ },
903
+ {
904
+ "epoch": 9.14,
905
+ "learning_rate": 4.3287160493827166e-05,
906
+ "loss": 0.0183,
907
+ "step": 12800
908
+ },
909
+ {
910
+ "epoch": 9.21,
911
+ "learning_rate": 4.3139259259259265e-05,
912
+ "loss": 0.0184,
913
+ "step": 12900
914
+ },
915
+ {
916
+ "epoch": 9.29,
917
+ "learning_rate": 4.2991358024691364e-05,
918
+ "loss": 0.0168,
919
+ "step": 13000
920
+ },
921
+ {
922
+ "epoch": 9.29,
923
+ "eval_iou": 0.982087265627057,
924
+ "step": 13000
925
+ },
926
+ {
927
+ "epoch": 9.36,
928
+ "learning_rate": 4.284345679012346e-05,
929
+ "loss": 0.0186,
930
+ "step": 13100
931
+ },
932
+ {
933
+ "epoch": 9.43,
934
+ "learning_rate": 4.2695555555555556e-05,
935
+ "loss": 0.0167,
936
+ "step": 13200
937
+ },
938
+ {
939
+ "epoch": 9.5,
940
+ "learning_rate": 4.2547654320987655e-05,
941
+ "loss": 0.0192,
942
+ "step": 13300
943
+ },
944
+ {
945
+ "epoch": 9.57,
946
+ "learning_rate": 4.2399753086419754e-05,
947
+ "loss": 0.017,
948
+ "step": 13400
949
+ },
950
+ {
951
+ "epoch": 9.64,
952
+ "learning_rate": 4.2251851851851854e-05,
953
+ "loss": 0.0174,
954
+ "step": 13500
955
+ },
956
+ {
957
+ "epoch": 9.64,
958
+ "eval_iou": 0.9819462275972554,
959
+ "step": 13500
960
+ },
961
+ {
962
+ "epoch": 9.71,
963
+ "learning_rate": 4.210395061728395e-05,
964
+ "loss": 0.017,
965
+ "step": 13600
966
+ },
967
+ {
968
+ "epoch": 9.79,
969
+ "learning_rate": 4.195604938271605e-05,
970
+ "loss": 0.0182,
971
+ "step": 13700
972
+ },
973
+ {
974
+ "epoch": 9.86,
975
+ "learning_rate": 4.180814814814815e-05,
976
+ "loss": 0.0179,
977
+ "step": 13800
978
+ },
979
+ {
980
+ "epoch": 9.93,
981
+ "learning_rate": 4.1660246913580244e-05,
982
+ "loss": 0.018,
983
+ "step": 13900
984
+ },
985
+ {
986
+ "epoch": 10.0,
987
+ "learning_rate": 4.151234567901234e-05,
988
+ "loss": 0.0179,
989
+ "step": 14000
990
+ },
991
+ {
992
+ "epoch": 10.0,
993
+ "eval_iou": 0.9823968813699835,
994
+ "step": 14000
995
+ },
996
+ {
997
+ "epoch": 10.0,
998
+ "eval_iou": 0.9823968813699835,
999
+ "step": 14000
1000
+ },
1001
+ {
1002
+ "epoch": 10.07,
1003
+ "learning_rate": 4.136444444444445e-05,
1004
+ "loss": 0.0176,
1005
+ "step": 14100
1006
+ },
1007
+ {
1008
+ "epoch": 10.14,
1009
+ "learning_rate": 4.121654320987655e-05,
1010
+ "loss": 0.0171,
1011
+ "step": 14200
1012
+ },
1013
+ {
1014
+ "epoch": 10.21,
1015
+ "learning_rate": 4.106864197530865e-05,
1016
+ "loss": 0.0178,
1017
+ "step": 14300
1018
+ },
1019
+ {
1020
+ "epoch": 10.29,
1021
+ "learning_rate": 4.092074074074074e-05,
1022
+ "loss": 0.0181,
1023
+ "step": 14400
1024
+ },
1025
+ {
1026
+ "epoch": 10.36,
1027
+ "learning_rate": 4.077283950617284e-05,
1028
+ "loss": 0.0187,
1029
+ "step": 14500
1030
+ },
1031
+ {
1032
+ "epoch": 10.36,
1033
+ "eval_iou": 0.9823005129309261,
1034
+ "step": 14500
1035
+ },
1036
+ {
1037
+ "epoch": 10.43,
1038
+ "learning_rate": 4.062493827160494e-05,
1039
+ "loss": 0.0189,
1040
+ "step": 14600
1041
+ },
1042
+ {
1043
+ "epoch": 10.5,
1044
+ "learning_rate": 4.047703703703704e-05,
1045
+ "loss": 0.0172,
1046
+ "step": 14700
1047
+ },
1048
+ {
1049
+ "epoch": 10.57,
1050
+ "learning_rate": 4.032913580246914e-05,
1051
+ "loss": 0.0175,
1052
+ "step": 14800
1053
+ },
1054
+ {
1055
+ "epoch": 10.64,
1056
+ "learning_rate": 4.0181234567901236e-05,
1057
+ "loss": 0.0179,
1058
+ "step": 14900
1059
+ },
1060
+ {
1061
+ "epoch": 10.71,
1062
+ "learning_rate": 4.003333333333334e-05,
1063
+ "loss": 0.0167,
1064
+ "step": 15000
1065
+ },
1066
+ {
1067
+ "epoch": 10.71,
1068
+ "eval_iou": 0.9822724671924815,
1069
+ "step": 15000
1070
+ },
1071
+ {
1072
+ "epoch": 10.79,
1073
+ "learning_rate": 3.9885432098765434e-05,
1074
+ "loss": 0.0175,
1075
+ "step": 15100
1076
+ },
1077
+ {
1078
+ "epoch": 10.86,
1079
+ "learning_rate": 3.973753086419753e-05,
1080
+ "loss": 0.0177,
1081
+ "step": 15200
1082
+ },
1083
+ {
1084
+ "epoch": 10.93,
1085
+ "learning_rate": 3.9589629629629626e-05,
1086
+ "loss": 0.0184,
1087
+ "step": 15300
1088
+ },
1089
+ {
1090
+ "epoch": 11.0,
1091
+ "learning_rate": 3.944172839506173e-05,
1092
+ "loss": 0.0174,
1093
+ "step": 15400
1094
+ },
1095
+ {
1096
+ "epoch": 11.07,
1097
+ "learning_rate": 3.929382716049383e-05,
1098
+ "loss": 0.0174,
1099
+ "step": 15500
1100
+ },
1101
+ {
1102
+ "epoch": 11.07,
1103
+ "eval_iou": 0.982211804857441,
1104
+ "step": 15500
1105
+ },
1106
+ {
1107
+ "epoch": 11.14,
1108
+ "learning_rate": 3.914592592592593e-05,
1109
+ "loss": 0.0168,
1110
+ "step": 15600
1111
+ },
1112
+ {
1113
+ "epoch": 11.21,
1114
+ "learning_rate": 3.899802469135802e-05,
1115
+ "loss": 0.0168,
1116
+ "step": 15700
1117
+ },
1118
+ {
1119
+ "epoch": 11.29,
1120
+ "learning_rate": 3.885012345679013e-05,
1121
+ "loss": 0.0173,
1122
+ "step": 15800
1123
+ },
1124
+ {
1125
+ "epoch": 11.36,
1126
+ "learning_rate": 3.870222222222222e-05,
1127
+ "loss": 0.0174,
1128
+ "step": 15900
1129
+ },
1130
+ {
1131
+ "epoch": 11.43,
1132
+ "learning_rate": 3.855432098765432e-05,
1133
+ "loss": 0.0179,
1134
+ "step": 16000
1135
+ },
1136
+ {
1137
+ "epoch": 11.43,
1138
+ "eval_iou": 0.9822788746917949,
1139
+ "step": 16000
1140
+ },
1141
+ {
1142
+ "epoch": 11.5,
1143
+ "learning_rate": 3.840641975308642e-05,
1144
+ "loss": 0.0178,
1145
+ "step": 16100
1146
+ },
1147
+ {
1148
+ "epoch": 11.57,
1149
+ "learning_rate": 3.825851851851852e-05,
1150
+ "loss": 0.0176,
1151
+ "step": 16200
1152
+ },
1153
+ {
1154
+ "epoch": 11.64,
1155
+ "learning_rate": 3.8110617283950625e-05,
1156
+ "loss": 0.0167,
1157
+ "step": 16300
1158
+ },
1159
+ {
1160
+ "epoch": 11.71,
1161
+ "learning_rate": 3.796271604938272e-05,
1162
+ "loss": 0.0175,
1163
+ "step": 16400
1164
+ },
1165
+ {
1166
+ "epoch": 11.79,
1167
+ "learning_rate": 3.781481481481482e-05,
1168
+ "loss": 0.0187,
1169
+ "step": 16500
1170
+ },
1171
+ {
1172
+ "epoch": 11.79,
1173
+ "eval_iou": 0.9822942889204213,
1174
+ "step": 16500
1175
+ },
1176
+ {
1177
+ "epoch": 11.86,
1178
+ "learning_rate": 3.766691358024691e-05,
1179
+ "loss": 0.0189,
1180
+ "step": 16600
1181
+ },
1182
+ {
1183
+ "epoch": 11.93,
1184
+ "learning_rate": 3.7519012345679015e-05,
1185
+ "loss": 0.0174,
1186
+ "step": 16700
1187
+ },
1188
+ {
1189
+ "epoch": 12.0,
1190
+ "learning_rate": 3.7371111111111114e-05,
1191
+ "loss": 0.0188,
1192
+ "step": 16800
1193
+ },
1194
+ {
1195
+ "epoch": 12.07,
1196
+ "learning_rate": 3.7223209876543214e-05,
1197
+ "loss": 0.0186,
1198
+ "step": 16900
1199
+ },
1200
+ {
1201
+ "epoch": 12.14,
1202
+ "learning_rate": 3.707530864197531e-05,
1203
+ "loss": 0.0184,
1204
+ "step": 17000
1205
+ },
1206
+ {
1207
+ "epoch": 12.14,
1208
+ "eval_iou": 0.9820192654927572,
1209
+ "step": 17000
1210
+ },
1211
+ {
1212
+ "epoch": 12.21,
1213
+ "learning_rate": 3.692740740740741e-05,
1214
+ "loss": 0.0167,
1215
+ "step": 17100
1216
+ },
1217
+ {
1218
+ "epoch": 12.29,
1219
+ "learning_rate": 3.6779506172839504e-05,
1220
+ "loss": 0.017,
1221
+ "step": 17200
1222
+ },
1223
+ {
1224
+ "epoch": 12.36,
1225
+ "learning_rate": 3.6631604938271604e-05,
1226
+ "loss": 0.0175,
1227
+ "step": 17300
1228
+ },
1229
+ {
1230
+ "epoch": 12.43,
1231
+ "learning_rate": 3.64837037037037e-05,
1232
+ "loss": 0.0179,
1233
+ "step": 17400
1234
+ },
1235
+ {
1236
+ "epoch": 12.5,
1237
+ "learning_rate": 3.633580246913581e-05,
1238
+ "loss": 0.0178,
1239
+ "step": 17500
1240
+ },
1241
+ {
1242
+ "epoch": 12.5,
1243
+ "eval_iou": 0.981830325196771,
1244
+ "step": 17500
1245
+ },
1246
+ {
1247
+ "epoch": 12.57,
1248
+ "learning_rate": 3.618790123456791e-05,
1249
+ "loss": 0.0179,
1250
+ "step": 17600
1251
+ },
1252
+ {
1253
+ "epoch": 12.64,
1254
+ "learning_rate": 3.604e-05,
1255
+ "loss": 0.0179,
1256
+ "step": 17700
1257
+ },
1258
+ {
1259
+ "epoch": 12.71,
1260
+ "learning_rate": 3.58920987654321e-05,
1261
+ "loss": 0.0172,
1262
+ "step": 17800
1263
+ },
1264
+ {
1265
+ "epoch": 12.79,
1266
+ "learning_rate": 3.574419753086419e-05,
1267
+ "loss": 0.017,
1268
+ "step": 17900
1269
+ },
1270
+ {
1271
+ "epoch": 12.86,
1272
+ "learning_rate": 3.55962962962963e-05,
1273
+ "loss": 0.0174,
1274
+ "step": 18000
1275
+ },
1276
+ {
1277
+ "epoch": 12.86,
1278
+ "eval_iou": 0.9824123674747991,
1279
+ "step": 18000
1280
+ },
1281
+ {
1282
+ "epoch": 12.93,
1283
+ "learning_rate": 3.54483950617284e-05,
1284
+ "loss": 0.0184,
1285
+ "step": 18100
1286
+ },
1287
+ {
1288
+ "epoch": 13.0,
1289
+ "learning_rate": 3.53004938271605e-05,
1290
+ "loss": 0.0172,
1291
+ "step": 18200
1292
+ },
1293
+ {
1294
+ "epoch": 13.07,
1295
+ "learning_rate": 3.5152592592592596e-05,
1296
+ "loss": 0.0171,
1297
+ "step": 18300
1298
+ },
1299
+ {
1300
+ "epoch": 13.14,
1301
+ "learning_rate": 3.5004691358024695e-05,
1302
+ "loss": 0.0172,
1303
+ "step": 18400
1304
+ },
1305
+ {
1306
+ "epoch": 13.21,
1307
+ "learning_rate": 3.4856790123456794e-05,
1308
+ "loss": 0.0173,
1309
+ "step": 18500
1310
+ },
1311
+ {
1312
+ "epoch": 13.21,
1313
+ "eval_iou": 0.9818773731297138,
1314
+ "step": 18500
1315
+ },
1316
+ {
1317
+ "epoch": 13.29,
1318
+ "learning_rate": 3.470888888888889e-05,
1319
+ "loss": 0.0167,
1320
+ "step": 18600
1321
+ },
1322
+ {
1323
+ "epoch": 13.36,
1324
+ "learning_rate": 3.4560987654320986e-05,
1325
+ "loss": 0.0168,
1326
+ "step": 18700
1327
+ },
1328
+ {
1329
+ "epoch": 13.43,
1330
+ "learning_rate": 3.441308641975309e-05,
1331
+ "loss": 0.0171,
1332
+ "step": 18800
1333
+ },
1334
+ {
1335
+ "epoch": 13.5,
1336
+ "learning_rate": 3.426518518518519e-05,
1337
+ "loss": 0.0165,
1338
+ "step": 18900
1339
+ },
1340
+ {
1341
+ "epoch": 13.57,
1342
+ "learning_rate": 3.411728395061729e-05,
1343
+ "loss": 0.0172,
1344
+ "step": 19000
1345
+ },
1346
+ {
1347
+ "epoch": 13.57,
1348
+ "eval_iou": 0.9824096741629582,
1349
+ "step": 19000
1350
+ },
1351
+ {
1352
+ "epoch": 13.64,
1353
+ "learning_rate": 3.396938271604938e-05,
1354
+ "loss": 0.0168,
1355
+ "step": 19100
1356
+ },
1357
+ {
1358
+ "epoch": 13.71,
1359
+ "learning_rate": 3.382148148148148e-05,
1360
+ "loss": 0.0172,
1361
+ "step": 19200
1362
+ },
1363
+ {
1364
+ "epoch": 13.79,
1365
+ "learning_rate": 3.367358024691358e-05,
1366
+ "loss": 0.017,
1367
+ "step": 19300
1368
+ },
1369
+ {
1370
+ "epoch": 13.86,
1371
+ "learning_rate": 3.352567901234568e-05,
1372
+ "loss": 0.0172,
1373
+ "step": 19400
1374
+ },
1375
+ {
1376
+ "epoch": 13.93,
1377
+ "learning_rate": 3.337777777777778e-05,
1378
+ "loss": 0.018,
1379
+ "step": 19500
1380
+ },
1381
+ {
1382
+ "epoch": 13.93,
1383
+ "eval_iou": 0.9822233032946494,
1384
+ "step": 19500
1385
+ },
1386
+ {
1387
+ "epoch": 14.0,
1388
+ "learning_rate": 3.322987654320988e-05,
1389
+ "loss": 0.017,
1390
+ "step": 19600
1391
+ },
1392
+ {
1393
+ "epoch": 14.07,
1394
+ "learning_rate": 3.308197530864198e-05,
1395
+ "loss": 0.017,
1396
+ "step": 19700
1397
+ },
1398
+ {
1399
+ "epoch": 14.14,
1400
+ "learning_rate": 3.293407407407408e-05,
1401
+ "loss": 0.0176,
1402
+ "step": 19800
1403
+ },
1404
+ {
1405
+ "epoch": 14.21,
1406
+ "learning_rate": 3.278617283950617e-05,
1407
+ "loss": 0.0173,
1408
+ "step": 19900
1409
+ },
1410
+ {
1411
+ "epoch": 14.29,
1412
+ "learning_rate": 3.263827160493827e-05,
1413
+ "loss": 0.0172,
1414
+ "step": 20000
1415
+ },
1416
+ {
1417
+ "epoch": 14.29,
1418
+ "eval_iou": 0.9821930264725405,
1419
+ "step": 20000
1420
+ },
1421
+ {
1422
+ "epoch": 14.36,
1423
+ "learning_rate": 3.2490370370370375e-05,
1424
+ "loss": 0.0174,
1425
+ "step": 20100
1426
+ },
1427
+ {
1428
+ "epoch": 14.43,
1429
+ "learning_rate": 3.2342469135802474e-05,
1430
+ "loss": 0.017,
1431
+ "step": 20200
1432
+ },
1433
+ {
1434
+ "epoch": 14.5,
1435
+ "learning_rate": 3.2194567901234573e-05,
1436
+ "loss": 0.0175,
1437
+ "step": 20300
1438
+ },
1439
+ {
1440
+ "epoch": 14.57,
1441
+ "learning_rate": 3.2046666666666666e-05,
1442
+ "loss": 0.0179,
1443
+ "step": 20400
1444
+ },
1445
+ {
1446
+ "epoch": 14.64,
1447
+ "learning_rate": 3.189876543209877e-05,
1448
+ "loss": 0.0173,
1449
+ "step": 20500
1450
+ },
1451
+ {
1452
+ "epoch": 14.64,
1453
+ "eval_iou": 0.9825124956813513,
1454
+ "step": 20500
1455
+ },
1456
+ {
1457
+ "epoch": 14.71,
1458
+ "learning_rate": 3.1750864197530864e-05,
1459
+ "loss": 0.0168,
1460
+ "step": 20600
1461
+ },
1462
+ {
1463
+ "epoch": 14.79,
1464
+ "learning_rate": 3.1602962962962964e-05,
1465
+ "loss": 0.018,
1466
+ "step": 20700
1467
+ },
1468
+ {
1469
+ "epoch": 14.86,
1470
+ "learning_rate": 3.145506172839506e-05,
1471
+ "loss": 0.0158,
1472
+ "step": 20800
1473
+ },
1474
+ {
1475
+ "epoch": 14.93,
1476
+ "learning_rate": 3.130716049382716e-05,
1477
+ "loss": 0.0174,
1478
+ "step": 20900
1479
+ },
1480
+ {
1481
+ "epoch": 15.0,
1482
+ "learning_rate": 3.115925925925927e-05,
1483
+ "loss": 0.017,
1484
+ "step": 21000
1485
+ },
1486
+ {
1487
+ "epoch": 15.0,
1488
+ "eval_iou": 0.9824787465965047,
1489
+ "step": 21000
1490
+ },
1491
+ {
1492
+ "epoch": 15.0,
1493
+ "eval_iou": 0.9824787465965047,
1494
+ "step": 21000
1495
+ },
1496
+ {
1497
+ "epoch": 15.07,
1498
+ "learning_rate": 3.101135802469136e-05,
1499
+ "loss": 0.0164,
1500
+ "step": 21100
1501
+ },
1502
+ {
1503
+ "epoch": 15.14,
1504
+ "learning_rate": 3.086345679012345e-05,
1505
+ "loss": 0.0169,
1506
+ "step": 21200
1507
+ },
1508
+ {
1509
+ "epoch": 15.21,
1510
+ "learning_rate": 3.071555555555555e-05,
1511
+ "loss": 0.0172,
1512
+ "step": 21300
1513
+ },
1514
+ {
1515
+ "epoch": 15.29,
1516
+ "learning_rate": 3.056765432098765e-05,
1517
+ "loss": 0.0171,
1518
+ "step": 21400
1519
+ },
1520
+ {
1521
+ "epoch": 15.36,
1522
+ "learning_rate": 3.0419753086419754e-05,
1523
+ "loss": 0.0174,
1524
+ "step": 21500
1525
+ },
1526
+ {
1527
+ "epoch": 15.36,
1528
+ "eval_iou": 0.9827606730601367,
1529
+ "step": 21500
1530
+ }
1531
+ ],
1532
+ "logging_steps": 100,
1533
+ "max_steps": 42000,
1534
+ "num_input_tokens_seen": 0,
1535
+ "num_train_epochs": 30,
1536
+ "save_steps": 500,
1537
+ "total_flos": 0.0,
1538
+ "train_batch_size": 10,
1539
+ "trial_name": null,
1540
+ "trial_params": null
1541
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4c15825d577677be79885136c7e48cda91e763d7c3ba8271106dcb6d3dd2179b
3
+ size 4728