Pham Thanh Dat commited on
Commit
de69c99
·
1 Parent(s): 5dc6355

Upload checkpoints

Browse files
checkpoint-78370/config.json ADDED
@@ -0,0 +1,67 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "datamonster/ms-cond-detr-res-50-vehicles",
3
+ "activation_dropout": 0.0,
4
+ "activation_function": "relu",
5
+ "architectures": [
6
+ "ConditionalDetrForObjectDetection"
7
+ ],
8
+ "attention_dropout": 0.0,
9
+ "auxiliary_loss": false,
10
+ "backbone": "resnet50",
11
+ "backbone_config": null,
12
+ "backbone_kwargs": {
13
+ "in_chans": 3,
14
+ "out_indices": [
15
+ 1,
16
+ 2,
17
+ 3,
18
+ 4
19
+ ]
20
+ },
21
+ "bbox_cost": 5,
22
+ "bbox_loss_coefficient": 5,
23
+ "class_cost": 2,
24
+ "cls_loss_coefficient": 2,
25
+ "d_model": 256,
26
+ "decoder_attention_heads": 8,
27
+ "decoder_ffn_dim": 2048,
28
+ "decoder_layerdrop": 0.0,
29
+ "decoder_layers": 6,
30
+ "dice_loss_coefficient": 1,
31
+ "dilation": false,
32
+ "dropout": 0.1,
33
+ "encoder_attention_heads": 8,
34
+ "encoder_ffn_dim": 2048,
35
+ "encoder_layerdrop": 0.0,
36
+ "encoder_layers": 6,
37
+ "focal_alpha": 0.25,
38
+ "giou_cost": 2,
39
+ "giou_loss_coefficient": 2,
40
+ "id2label": {
41
+ "0": "motorbike",
42
+ "1": "car",
43
+ "2": "bus",
44
+ "3": "container"
45
+ },
46
+ "init_std": 0.02,
47
+ "init_xavier_std": 1.0,
48
+ "is_encoder_decoder": true,
49
+ "label2id": {
50
+ "bus": 2,
51
+ "car": 1,
52
+ "container": 3,
53
+ "motorbike": 0
54
+ },
55
+ "mask_loss_coefficient": 1,
56
+ "max_position_embeddings": 1024,
57
+ "model_type": "conditional_detr",
58
+ "num_channels": 3,
59
+ "num_hidden_layers": 6,
60
+ "num_queries": 300,
61
+ "position_embedding_type": "sine",
62
+ "scale_embedding": false,
63
+ "torch_dtype": "float32",
64
+ "transformers_version": "4.45.1",
65
+ "use_pretrained_backbone": true,
66
+ "use_timm_backbone": true
67
+ }
checkpoint-78370/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ce8fafc752a7af82ca7b4e019ebf7c84b6b9176bd3395b346b322247ef2276ad
3
+ size 174078768
checkpoint-78370/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:722b7433810d5b751cb6c7b15eb4643a962f3aaef5d856648770c820fa9a712f
3
+ size 187563450
checkpoint-78370/preprocessor_config.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_convert_annotations": true,
3
+ "do_normalize": true,
4
+ "do_pad": true,
5
+ "do_rescale": true,
6
+ "do_resize": true,
7
+ "format": "coco_detection",
8
+ "image_mean": [
9
+ 0.485,
10
+ 0.456,
11
+ 0.406
12
+ ],
13
+ "image_processor_type": "ConditionalDetrImageProcessor",
14
+ "image_std": [
15
+ 0.229,
16
+ 0.224,
17
+ 0.225
18
+ ],
19
+ "pad_size": {
20
+ "height": 768,
21
+ "width": 768
22
+ },
23
+ "resample": 2,
24
+ "rescale_factor": 0.00392156862745098,
25
+ "size": {
26
+ "max_height": 768,
27
+ "max_width": 768
28
+ }
29
+ }
checkpoint-78370/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f7b42cb377b2b7a3fe5708652091af2e8accb72137e89f3630b8b485458631eb
3
+ size 14244
checkpoint-78370/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f5b570eb2601477f0617c707c2fbe0d15e27438170f7908770be0a4898718e3c
3
+ size 1000
checkpoint-78370/trainer_state.json ADDED
@@ -0,0 +1,2077 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.5035,
3
+ "best_model_checkpoint": "ms-cond-detr-res-50-vehicles/checkpoint-73760",
4
+ "epoch": 34.0,
5
+ "eval_steps": 500,
6
+ "global_step": 78370,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.21691973969631237,
13
+ "grad_norm": 38753.67578125,
14
+ "learning_rate": 9.999686465963656e-05,
15
+ "loss": 1.2341,
16
+ "step": 500
17
+ },
18
+ {
19
+ "epoch": 0.43383947939262474,
20
+ "grad_norm": 14156.765625,
21
+ "learning_rate": 9.998728034076959e-05,
22
+ "loss": 1.1441,
23
+ "step": 1000
24
+ },
25
+ {
26
+ "epoch": 0.6507592190889371,
27
+ "grad_norm": 7688.85986328125,
28
+ "learning_rate": 9.99712861472983e-05,
29
+ "loss": 1.1357,
30
+ "step": 1500
31
+ },
32
+ {
33
+ "epoch": 0.8676789587852495,
34
+ "grad_norm": 10963.5341796875,
35
+ "learning_rate": 9.994882003818517e-05,
36
+ "loss": 1.1322,
37
+ "step": 2000
38
+ },
39
+ {
40
+ "epoch": 1.0,
41
+ "eval_loss": 1.0705492496490479,
42
+ "eval_map": 0.3886,
43
+ "eval_map_50": 0.7276,
44
+ "eval_map_75": 0.371,
45
+ "eval_map_bus": 0.4553,
46
+ "eval_map_car": 0.4267,
47
+ "eval_map_container": 0.4051,
48
+ "eval_map_large": 0.6092,
49
+ "eval_map_medium": 0.3725,
50
+ "eval_map_motorbike": 0.2674,
51
+ "eval_map_small": 0.0755,
52
+ "eval_mar_1": 0.2228,
53
+ "eval_mar_10": 0.4736,
54
+ "eval_mar_100": 0.5226,
55
+ "eval_mar_100_bus": 0.6014,
56
+ "eval_mar_100_car": 0.5398,
57
+ "eval_mar_100_container": 0.5541,
58
+ "eval_mar_100_motorbike": 0.395,
59
+ "eval_mar_large": 0.7004,
60
+ "eval_mar_medium": 0.5176,
61
+ "eval_mar_small": 0.2475,
62
+ "eval_runtime": 149.8641,
63
+ "eval_samples_per_second": 15.367,
64
+ "eval_steps_per_second": 1.922,
65
+ "step": 2305
66
+ },
67
+ {
68
+ "epoch": 1.0845986984815619,
69
+ "grad_norm": 8778.8525390625,
70
+ "learning_rate": 9.991991051623087e-05,
71
+ "loss": 1.1204,
72
+ "step": 2500
73
+ },
74
+ {
75
+ "epoch": 1.3015184381778742,
76
+ "grad_norm": 8215.1748046875,
77
+ "learning_rate": 9.988456131077249e-05,
78
+ "loss": 1.0925,
79
+ "step": 3000
80
+ },
81
+ {
82
+ "epoch": 1.5184381778741867,
83
+ "grad_norm": 15722.8212890625,
84
+ "learning_rate": 9.984277698186814e-05,
85
+ "loss": 1.0867,
86
+ "step": 3500
87
+ },
88
+ {
89
+ "epoch": 1.735357917570499,
90
+ "grad_norm": 17670.681640625,
91
+ "learning_rate": 9.979456291970887e-05,
92
+ "loss": 1.0714,
93
+ "step": 4000
94
+ },
95
+ {
96
+ "epoch": 1.9522776572668112,
97
+ "grad_norm": 15503.669921875,
98
+ "learning_rate": 9.974004102519452e-05,
99
+ "loss": 1.0605,
100
+ "step": 4500
101
+ },
102
+ {
103
+ "epoch": 2.0,
104
+ "eval_loss": 1.0418221950531006,
105
+ "eval_map": 0.3977,
106
+ "eval_map_50": 0.7371,
107
+ "eval_map_75": 0.389,
108
+ "eval_map_bus": 0.4529,
109
+ "eval_map_car": 0.4318,
110
+ "eval_map_container": 0.4316,
111
+ "eval_map_large": 0.6212,
112
+ "eval_map_medium": 0.3842,
113
+ "eval_map_motorbike": 0.2745,
114
+ "eval_map_small": 0.0841,
115
+ "eval_mar_1": 0.2265,
116
+ "eval_mar_10": 0.4923,
117
+ "eval_mar_100": 0.5393,
118
+ "eval_mar_100_bus": 0.6241,
119
+ "eval_mar_100_car": 0.5529,
120
+ "eval_mar_100_container": 0.5848,
121
+ "eval_mar_100_motorbike": 0.3954,
122
+ "eval_mar_large": 0.7101,
123
+ "eval_mar_medium": 0.5389,
124
+ "eval_mar_small": 0.2511,
125
+ "eval_runtime": 146.5587,
126
+ "eval_samples_per_second": 15.714,
127
+ "eval_steps_per_second": 1.965,
128
+ "step": 4610
129
+ },
130
+ {
131
+ "epoch": 2.1691973969631237,
132
+ "grad_norm": 3109.66162109375,
133
+ "learning_rate": 9.967899980939251e-05,
134
+ "loss": 1.0591,
135
+ "step": 5000
136
+ },
137
+ {
138
+ "epoch": 2.3861171366594363,
139
+ "grad_norm": 5812.068359375,
140
+ "learning_rate": 9.961154998764017e-05,
141
+ "loss": 1.0676,
142
+ "step": 5500
143
+ },
144
+ {
145
+ "epoch": 2.6030368763557483,
146
+ "grad_norm": 7711.87158203125,
147
+ "learning_rate": 9.953770026098485e-05,
148
+ "loss": 1.0673,
149
+ "step": 6000
150
+ },
151
+ {
152
+ "epoch": 2.819956616052061,
153
+ "grad_norm": 11271.8916015625,
154
+ "learning_rate": 9.945746015606358e-05,
155
+ "loss": 1.0634,
156
+ "step": 6500
157
+ },
158
+ {
159
+ "epoch": 3.0,
160
+ "eval_loss": 1.0088645219802856,
161
+ "eval_map": 0.4231,
162
+ "eval_map_50": 0.758,
163
+ "eval_map_75": 0.4223,
164
+ "eval_map_bus": 0.4788,
165
+ "eval_map_car": 0.4565,
166
+ "eval_map_container": 0.4648,
167
+ "eval_map_large": 0.658,
168
+ "eval_map_medium": 0.4084,
169
+ "eval_map_motorbike": 0.2923,
170
+ "eval_map_small": 0.09,
171
+ "eval_mar_1": 0.2352,
172
+ "eval_mar_10": 0.5048,
173
+ "eval_mar_100": 0.555,
174
+ "eval_mar_100_bus": 0.6323,
175
+ "eval_mar_100_car": 0.5672,
176
+ "eval_mar_100_container": 0.6028,
177
+ "eval_mar_100_motorbike": 0.4178,
178
+ "eval_mar_large": 0.7546,
179
+ "eval_mar_medium": 0.5521,
180
+ "eval_mar_small": 0.2802,
181
+ "eval_runtime": 147.4693,
182
+ "eval_samples_per_second": 15.617,
183
+ "eval_steps_per_second": 1.953,
184
+ "step": 6915
185
+ },
186
+ {
187
+ "epoch": 3.036876355748373,
188
+ "grad_norm": 20180.71484375,
189
+ "learning_rate": 9.937084002387407e-05,
190
+ "loss": 1.0418,
191
+ "step": 7000
192
+ },
193
+ {
194
+ "epoch": 3.2537960954446854,
195
+ "grad_norm": 11818.376953125,
196
+ "learning_rate": 9.927804336468942e-05,
197
+ "loss": 1.0386,
198
+ "step": 7500
199
+ },
200
+ {
201
+ "epoch": 3.470715835140998,
202
+ "grad_norm": 9070.5185546875,
203
+ "learning_rate": 9.917871022281522e-05,
204
+ "loss": 1.0635,
205
+ "step": 8000
206
+ },
207
+ {
208
+ "epoch": 3.6876355748373104,
209
+ "grad_norm": 7446.32666015625,
210
+ "learning_rate": 9.907303301249819e-05,
211
+ "loss": 1.0466,
212
+ "step": 8500
213
+ },
214
+ {
215
+ "epoch": 3.9045553145336225,
216
+ "grad_norm": 2428.818603515625,
217
+ "learning_rate": 9.896102536613032e-05,
218
+ "loss": 1.0288,
219
+ "step": 9000
220
+ },
221
+ {
222
+ "epoch": 4.0,
223
+ "eval_loss": 1.002821922302246,
224
+ "eval_map": 0.4297,
225
+ "eval_map_50": 0.7647,
226
+ "eval_map_75": 0.429,
227
+ "eval_map_bus": 0.4925,
228
+ "eval_map_car": 0.4582,
229
+ "eval_map_container": 0.4685,
230
+ "eval_map_large": 0.6667,
231
+ "eval_map_medium": 0.4184,
232
+ "eval_map_motorbike": 0.2997,
233
+ "eval_map_small": 0.0816,
234
+ "eval_mar_1": 0.2383,
235
+ "eval_mar_10": 0.5069,
236
+ "eval_mar_100": 0.5549,
237
+ "eval_mar_100_bus": 0.634,
238
+ "eval_mar_100_car": 0.5649,
239
+ "eval_mar_100_container": 0.5965,
240
+ "eval_mar_100_motorbike": 0.4243,
241
+ "eval_mar_large": 0.7579,
242
+ "eval_mar_medium": 0.5516,
243
+ "eval_mar_small": 0.2596,
244
+ "eval_runtime": 146.3143,
245
+ "eval_samples_per_second": 15.74,
246
+ "eval_steps_per_second": 1.968,
247
+ "step": 9220
248
+ },
249
+ {
250
+ "epoch": 4.1214750542299345,
251
+ "grad_norm": 4431.853515625,
252
+ "learning_rate": 9.88429446733437e-05,
253
+ "loss": 1.0289,
254
+ "step": 9500
255
+ },
256
+ {
257
+ "epoch": 4.3383947939262475,
258
+ "grad_norm": 16948.7890625,
259
+ "learning_rate": 9.871833290237144e-05,
260
+ "loss": 1.031,
261
+ "step": 10000
262
+ },
263
+ {
264
+ "epoch": 4.55531453362256,
265
+ "grad_norm": 5605.24951171875,
266
+ "learning_rate": 9.858743645176947e-05,
267
+ "loss": 1.0303,
268
+ "step": 10500
269
+ },
270
+ {
271
+ "epoch": 4.7722342733188725,
272
+ "grad_norm": 16235.091796875,
273
+ "learning_rate": 9.845027220721935e-05,
274
+ "loss": 1.0206,
275
+ "step": 11000
276
+ },
277
+ {
278
+ "epoch": 4.989154013015185,
279
+ "grad_norm": 20350.36328125,
280
+ "learning_rate": 9.830685786294999e-05,
281
+ "loss": 1.0226,
282
+ "step": 11500
283
+ },
284
+ {
285
+ "epoch": 5.0,
286
+ "eval_loss": 0.9774508476257324,
287
+ "eval_map": 0.4366,
288
+ "eval_map_50": 0.7785,
289
+ "eval_map_75": 0.4388,
290
+ "eval_map_bus": 0.5023,
291
+ "eval_map_car": 0.4658,
292
+ "eval_map_container": 0.4766,
293
+ "eval_map_large": 0.6605,
294
+ "eval_map_medium": 0.4243,
295
+ "eval_map_motorbike": 0.3015,
296
+ "eval_map_small": 0.0942,
297
+ "eval_mar_1": 0.2402,
298
+ "eval_mar_10": 0.5173,
299
+ "eval_mar_100": 0.5666,
300
+ "eval_mar_100_bus": 0.6534,
301
+ "eval_mar_100_car": 0.5747,
302
+ "eval_mar_100_container": 0.6145,
303
+ "eval_mar_100_motorbike": 0.4239,
304
+ "eval_mar_large": 0.751,
305
+ "eval_mar_medium": 0.5659,
306
+ "eval_mar_small": 0.2757,
307
+ "eval_runtime": 146.1171,
308
+ "eval_samples_per_second": 15.761,
309
+ "eval_steps_per_second": 1.971,
310
+ "step": 11525
311
+ },
312
+ {
313
+ "epoch": 5.206073752711497,
314
+ "grad_norm": 11680.388671875,
315
+ "learning_rate": 9.815721191945502e-05,
316
+ "loss": 1.0021,
317
+ "step": 12000
318
+ },
319
+ {
320
+ "epoch": 5.422993492407809,
321
+ "grad_norm": 6468.74072265625,
322
+ "learning_rate": 9.800167158422275e-05,
323
+ "loss": 1.0049,
324
+ "step": 12500
325
+ },
326
+ {
327
+ "epoch": 5.639913232104122,
328
+ "grad_norm": 10755.380859375,
329
+ "learning_rate": 9.78396335205612e-05,
330
+ "loss": 1.016,
331
+ "step": 13000
332
+ },
333
+ {
334
+ "epoch": 5.856832971800434,
335
+ "grad_norm": 6588.42578125,
336
+ "learning_rate": 9.767142412975505e-05,
337
+ "loss": 1.022,
338
+ "step": 13500
339
+ },
340
+ {
341
+ "epoch": 6.0,
342
+ "eval_loss": 0.9850358366966248,
343
+ "eval_map": 0.4356,
344
+ "eval_map_50": 0.7775,
345
+ "eval_map_75": 0.4345,
346
+ "eval_map_bus": 0.5032,
347
+ "eval_map_car": 0.455,
348
+ "eval_map_container": 0.4783,
349
+ "eval_map_large": 0.661,
350
+ "eval_map_medium": 0.4213,
351
+ "eval_map_motorbike": 0.3059,
352
+ "eval_map_small": 0.106,
353
+ "eval_mar_1": 0.2402,
354
+ "eval_mar_10": 0.514,
355
+ "eval_mar_100": 0.563,
356
+ "eval_mar_100_bus": 0.651,
357
+ "eval_mar_100_car": 0.5596,
358
+ "eval_mar_100_container": 0.6163,
359
+ "eval_mar_100_motorbike": 0.4251,
360
+ "eval_mar_large": 0.7479,
361
+ "eval_mar_medium": 0.5634,
362
+ "eval_mar_small": 0.2541,
363
+ "eval_runtime": 145.7736,
364
+ "eval_samples_per_second": 15.798,
365
+ "eval_steps_per_second": 1.976,
366
+ "step": 13830
367
+ },
368
+ {
369
+ "epoch": 6.073752711496746,
370
+ "grad_norm": 8365.4345703125,
371
+ "learning_rate": 9.749706511086642e-05,
372
+ "loss": 1.0065,
373
+ "step": 14000
374
+ },
375
+ {
376
+ "epoch": 6.290672451193059,
377
+ "grad_norm": 7092.89306640625,
378
+ "learning_rate": 9.731694602810796e-05,
379
+ "loss": 1.0004,
380
+ "step": 14500
381
+ },
382
+ {
383
+ "epoch": 6.507592190889371,
384
+ "grad_norm": 7779.22314453125,
385
+ "learning_rate": 9.713036820450053e-05,
386
+ "loss": 1.0193,
387
+ "step": 15000
388
+ },
389
+ {
390
+ "epoch": 6.724511930585683,
391
+ "grad_norm": 5161.82958984375,
392
+ "learning_rate": 9.693771054918806e-05,
393
+ "loss": 0.9992,
394
+ "step": 15500
395
+ },
396
+ {
397
+ "epoch": 6.941431670281996,
398
+ "grad_norm": 5342.78662109375,
399
+ "learning_rate": 9.673899791506617e-05,
400
+ "loss": 0.9936,
401
+ "step": 16000
402
+ },
403
+ {
404
+ "epoch": 7.0,
405
+ "eval_loss": 0.9677829742431641,
406
+ "eval_map": 0.4394,
407
+ "eval_map_50": 0.781,
408
+ "eval_map_75": 0.4415,
409
+ "eval_map_bus": 0.5059,
410
+ "eval_map_car": 0.4716,
411
+ "eval_map_container": 0.4731,
412
+ "eval_map_large": 0.6636,
413
+ "eval_map_medium": 0.4308,
414
+ "eval_map_motorbike": 0.307,
415
+ "eval_map_small": 0.109,
416
+ "eval_mar_1": 0.2401,
417
+ "eval_mar_10": 0.5184,
418
+ "eval_mar_100": 0.5674,
419
+ "eval_mar_100_bus": 0.6546,
420
+ "eval_mar_100_car": 0.5736,
421
+ "eval_mar_100_container": 0.6126,
422
+ "eval_mar_100_motorbike": 0.4289,
423
+ "eval_mar_large": 0.7468,
424
+ "eval_mar_medium": 0.5696,
425
+ "eval_mar_small": 0.2775,
426
+ "eval_runtime": 147.1057,
427
+ "eval_samples_per_second": 15.655,
428
+ "eval_steps_per_second": 1.958,
429
+ "step": 16135
430
+ },
431
+ {
432
+ "epoch": 7.158351409978308,
433
+ "grad_norm": 17945.982421875,
434
+ "learning_rate": 9.653425593612456e-05,
435
+ "loss": 0.9817,
436
+ "step": 16500
437
+ },
438
+ {
439
+ "epoch": 7.375271149674621,
440
+ "grad_norm": 2631.81591796875,
441
+ "learning_rate": 9.632393848695017e-05,
442
+ "loss": 0.9845,
443
+ "step": 17000
444
+ },
445
+ {
446
+ "epoch": 7.592190889370933,
447
+ "grad_norm": 1825.7994384765625,
448
+ "learning_rate": 9.61072297519293e-05,
449
+ "loss": 0.9839,
450
+ "step": 17500
451
+ },
452
+ {
453
+ "epoch": 7.809110629067245,
454
+ "grad_norm": 8929.837890625,
455
+ "learning_rate": 9.588457317037219e-05,
456
+ "loss": 0.9908,
457
+ "step": 18000
458
+ },
459
+ {
460
+ "epoch": 8.0,
461
+ "eval_loss": 0.9544239640235901,
462
+ "eval_map": 0.4515,
463
+ "eval_map_50": 0.7929,
464
+ "eval_map_75": 0.459,
465
+ "eval_map_bus": 0.5215,
466
+ "eval_map_car": 0.4759,
467
+ "eval_map_container": 0.4956,
468
+ "eval_map_large": 0.6707,
469
+ "eval_map_medium": 0.4407,
470
+ "eval_map_motorbike": 0.3131,
471
+ "eval_map_small": 0.103,
472
+ "eval_mar_1": 0.2439,
473
+ "eval_mar_10": 0.5243,
474
+ "eval_mar_100": 0.573,
475
+ "eval_mar_100_bus": 0.6558,
476
+ "eval_mar_100_car": 0.579,
477
+ "eval_mar_100_container": 0.6217,
478
+ "eval_mar_100_motorbike": 0.4355,
479
+ "eval_mar_large": 0.7575,
480
+ "eval_mar_medium": 0.5708,
481
+ "eval_mar_small": 0.3107,
482
+ "eval_runtime": 147.1314,
483
+ "eval_samples_per_second": 15.653,
484
+ "eval_steps_per_second": 1.957,
485
+ "step": 18440
486
+ },
487
+ {
488
+ "epoch": 8.026030368763557,
489
+ "grad_norm": 9851.580078125,
490
+ "learning_rate": 9.565599746504495e-05,
491
+ "loss": 0.9914,
492
+ "step": 18500
493
+ },
494
+ {
495
+ "epoch": 8.242950108459869,
496
+ "grad_norm": 46131.0703125,
497
+ "learning_rate": 9.542153212228251e-05,
498
+ "loss": 0.9789,
499
+ "step": 19000
500
+ },
501
+ {
502
+ "epoch": 8.459869848156183,
503
+ "grad_norm": 40477.8671875,
504
+ "learning_rate": 9.518120738818473e-05,
505
+ "loss": 0.9642,
506
+ "step": 19500
507
+ },
508
+ {
509
+ "epoch": 8.676789587852495,
510
+ "grad_norm": 21795.466796875,
511
+ "learning_rate": 9.493505426471479e-05,
512
+ "loss": 0.9637,
513
+ "step": 20000
514
+ },
515
+ {
516
+ "epoch": 8.893709327548807,
517
+ "grad_norm": 5294.40673828125,
518
+ "learning_rate": 9.468361416878137e-05,
519
+ "loss": 0.9853,
520
+ "step": 20500
521
+ },
522
+ {
523
+ "epoch": 9.0,
524
+ "eval_loss": 0.9467804431915283,
525
+ "eval_map": 0.4529,
526
+ "eval_map_50": 0.7951,
527
+ "eval_map_75": 0.46,
528
+ "eval_map_bus": 0.5273,
529
+ "eval_map_car": 0.4728,
530
+ "eval_map_container": 0.4969,
531
+ "eval_map_large": 0.6735,
532
+ "eval_map_medium": 0.4421,
533
+ "eval_map_motorbike": 0.3144,
534
+ "eval_map_small": 0.1016,
535
+ "eval_mar_1": 0.2466,
536
+ "eval_mar_10": 0.5271,
537
+ "eval_mar_100": 0.5772,
538
+ "eval_mar_100_bus": 0.6592,
539
+ "eval_mar_100_car": 0.5831,
540
+ "eval_mar_100_container": 0.6281,
541
+ "eval_mar_100_motorbike": 0.4385,
542
+ "eval_mar_large": 0.7607,
543
+ "eval_mar_medium": 0.5767,
544
+ "eval_mar_small": 0.2947,
545
+ "eval_runtime": 146.7224,
546
+ "eval_samples_per_second": 15.696,
547
+ "eval_steps_per_second": 1.963,
548
+ "step": 20745
549
+ },
550
+ {
551
+ "epoch": 9.11062906724512,
552
+ "grad_norm": 20100.53515625,
553
+ "learning_rate": 9.442591177115253e-05,
554
+ "loss": 0.973,
555
+ "step": 21000
556
+ },
557
+ {
558
+ "epoch": 9.327548806941431,
559
+ "grad_norm": 4432.14404296875,
560
+ "learning_rate": 9.416247841751377e-05,
561
+ "loss": 0.9844,
562
+ "step": 21500
563
+ },
564
+ {
565
+ "epoch": 9.544468546637743,
566
+ "grad_norm": 5448.53369140625,
567
+ "learning_rate": 9.389334809084711e-05,
568
+ "loss": 0.9783,
569
+ "step": 22000
570
+ },
571
+ {
572
+ "epoch": 9.761388286334057,
573
+ "grad_norm": 18744.869140625,
574
+ "learning_rate": 9.361855550904582e-05,
575
+ "loss": 0.9591,
576
+ "step": 22500
577
+ },
578
+ {
579
+ "epoch": 9.97830802603037,
580
+ "grad_norm": 9256.87890625,
581
+ "learning_rate": 9.333870255084278e-05,
582
+ "loss": 0.9536,
583
+ "step": 23000
584
+ },
585
+ {
586
+ "epoch": 10.0,
587
+ "eval_loss": 0.9346117377281189,
588
+ "eval_map": 0.4603,
589
+ "eval_map_50": 0.8028,
590
+ "eval_map_75": 0.4763,
591
+ "eval_map_bus": 0.5317,
592
+ "eval_map_car": 0.4808,
593
+ "eval_map_container": 0.5064,
594
+ "eval_map_large": 0.6826,
595
+ "eval_map_medium": 0.4525,
596
+ "eval_map_motorbike": 0.3221,
597
+ "eval_map_small": 0.108,
598
+ "eval_mar_1": 0.2479,
599
+ "eval_mar_10": 0.532,
600
+ "eval_mar_100": 0.5823,
601
+ "eval_mar_100_bus": 0.6637,
602
+ "eval_mar_100_car": 0.591,
603
+ "eval_mar_100_container": 0.6308,
604
+ "eval_mar_100_motorbike": 0.4437,
605
+ "eval_mar_large": 0.7735,
606
+ "eval_mar_medium": 0.5844,
607
+ "eval_mar_small": 0.2885,
608
+ "eval_runtime": 144.7391,
609
+ "eval_samples_per_second": 15.911,
610
+ "eval_steps_per_second": 1.99,
611
+ "step": 23050
612
+ },
613
+ {
614
+ "epoch": 10.195227765726681,
615
+ "grad_norm": 6654.49365234375,
616
+ "learning_rate": 9.30532812271677e-05,
617
+ "loss": 0.9821,
618
+ "step": 23500
619
+ },
620
+ {
621
+ "epoch": 10.412147505422993,
622
+ "grad_norm": 12540.853515625,
623
+ "learning_rate": 9.276173960919273e-05,
624
+ "loss": 0.9966,
625
+ "step": 24000
626
+ },
627
+ {
628
+ "epoch": 10.629067245119305,
629
+ "grad_norm": 8583.029296875,
630
+ "learning_rate": 9.246468171391835e-05,
631
+ "loss": 0.9826,
632
+ "step": 24500
633
+ },
634
+ {
635
+ "epoch": 10.845986984815617,
636
+ "grad_norm": 5945.16796875,
637
+ "learning_rate": 9.216214586190261e-05,
638
+ "loss": 0.9769,
639
+ "step": 25000
640
+ },
641
+ {
642
+ "epoch": 11.0,
643
+ "eval_loss": 0.9336058497428894,
644
+ "eval_map": 0.4511,
645
+ "eval_map_50": 0.8,
646
+ "eval_map_75": 0.4575,
647
+ "eval_map_bus": 0.5073,
648
+ "eval_map_car": 0.4708,
649
+ "eval_map_container": 0.5006,
650
+ "eval_map_large": 0.6611,
651
+ "eval_map_medium": 0.4469,
652
+ "eval_map_motorbike": 0.3259,
653
+ "eval_map_small": 0.1072,
654
+ "eval_mar_1": 0.2412,
655
+ "eval_mar_10": 0.5266,
656
+ "eval_mar_100": 0.5752,
657
+ "eval_mar_100_bus": 0.6493,
658
+ "eval_mar_100_car": 0.5772,
659
+ "eval_mar_100_container": 0.6233,
660
+ "eval_mar_100_motorbike": 0.4511,
661
+ "eval_mar_large": 0.7508,
662
+ "eval_mar_medium": 0.5789,
663
+ "eval_mar_small": 0.2908,
664
+ "eval_runtime": 147.6126,
665
+ "eval_samples_per_second": 15.602,
666
+ "eval_steps_per_second": 1.951,
667
+ "step": 25355
668
+ },
669
+ {
670
+ "epoch": 11.062906724511931,
671
+ "grad_norm": 10052.8896484375,
672
+ "learning_rate": 9.185417108036163e-05,
673
+ "loss": 0.9731,
674
+ "step": 25500
675
+ },
676
+ {
677
+ "epoch": 11.279826464208243,
678
+ "grad_norm": 3725.647216796875,
679
+ "learning_rate": 9.154079709813505e-05,
680
+ "loss": 0.9572,
681
+ "step": 26000
682
+ },
683
+ {
684
+ "epoch": 11.496746203904555,
685
+ "grad_norm": 10829.8896484375,
686
+ "learning_rate": 9.122206434056104e-05,
687
+ "loss": 0.9558,
688
+ "step": 26500
689
+ },
690
+ {
691
+ "epoch": 11.713665943600867,
692
+ "grad_norm": 16212.732421875,
693
+ "learning_rate": 9.08980139242614e-05,
694
+ "loss": 0.9622,
695
+ "step": 27000
696
+ },
697
+ {
698
+ "epoch": 11.93058568329718,
699
+ "grad_norm": 12326.623046875,
700
+ "learning_rate": 9.056868765183752e-05,
701
+ "loss": 0.9506,
702
+ "step": 27500
703
+ },
704
+ {
705
+ "epoch": 12.0,
706
+ "eval_loss": 0.9213413596153259,
707
+ "eval_map": 0.4617,
708
+ "eval_map_50": 0.8095,
709
+ "eval_map_75": 0.4735,
710
+ "eval_map_bus": 0.5252,
711
+ "eval_map_car": 0.4891,
712
+ "eval_map_container": 0.5052,
713
+ "eval_map_large": 0.676,
714
+ "eval_map_medium": 0.455,
715
+ "eval_map_motorbike": 0.3272,
716
+ "eval_map_small": 0.1083,
717
+ "eval_mar_1": 0.2461,
718
+ "eval_mar_10": 0.5327,
719
+ "eval_mar_100": 0.5818,
720
+ "eval_mar_100_bus": 0.6626,
721
+ "eval_mar_100_car": 0.5902,
722
+ "eval_mar_100_container": 0.627,
723
+ "eval_mar_100_motorbike": 0.4475,
724
+ "eval_mar_large": 0.7631,
725
+ "eval_mar_medium": 0.5838,
726
+ "eval_mar_small": 0.2935,
727
+ "eval_runtime": 143.9851,
728
+ "eval_samples_per_second": 15.995,
729
+ "eval_steps_per_second": 2.0,
730
+ "step": 27660
731
+ },
732
+ {
733
+ "epoch": 12.147505422993492,
734
+ "grad_norm": 23455.04296875,
735
+ "learning_rate": 9.023412800647782e-05,
736
+ "loss": 0.9338,
737
+ "step": 28000
738
+ },
739
+ {
740
+ "epoch": 12.364425162689805,
741
+ "grad_norm": 7761.9912109375,
742
+ "learning_rate": 8.989506279701155e-05,
743
+ "loss": 0.9535,
744
+ "step": 28500
745
+ },
746
+ {
747
+ "epoch": 12.581344902386117,
748
+ "grad_norm": 7034.04052734375,
749
+ "learning_rate": 8.955017679879593e-05,
750
+ "loss": 0.9507,
751
+ "step": 29000
752
+ },
753
+ {
754
+ "epoch": 12.79826464208243,
755
+ "grad_norm": 5548.38525390625,
756
+ "learning_rate": 8.920018881585156e-05,
757
+ "loss": 0.9417,
758
+ "step": 29500
759
+ },
760
+ {
761
+ "epoch": 13.0,
762
+ "eval_loss": 0.9276334643363953,
763
+ "eval_map": 0.4608,
764
+ "eval_map_50": 0.8076,
765
+ "eval_map_75": 0.4687,
766
+ "eval_map_bus": 0.5336,
767
+ "eval_map_car": 0.4866,
768
+ "eval_map_container": 0.4992,
769
+ "eval_map_large": 0.6724,
770
+ "eval_map_medium": 0.4511,
771
+ "eval_map_motorbike": 0.3238,
772
+ "eval_map_small": 0.1064,
773
+ "eval_mar_1": 0.2472,
774
+ "eval_mar_10": 0.5288,
775
+ "eval_mar_100": 0.5752,
776
+ "eval_mar_100_bus": 0.6588,
777
+ "eval_mar_100_car": 0.5895,
778
+ "eval_mar_100_container": 0.6158,
779
+ "eval_mar_100_motorbike": 0.4369,
780
+ "eval_mar_large": 0.7512,
781
+ "eval_mar_medium": 0.5708,
782
+ "eval_mar_small": 0.2981,
783
+ "eval_runtime": 145.1236,
784
+ "eval_samples_per_second": 15.869,
785
+ "eval_steps_per_second": 1.985,
786
+ "step": 29965
787
+ },
788
+ {
789
+ "epoch": 13.015184381778742,
790
+ "grad_norm": 12253.849609375,
791
+ "learning_rate": 8.884514399673369e-05,
792
+ "loss": 0.9555,
793
+ "step": 30000
794
+ },
795
+ {
796
+ "epoch": 13.232104121475054,
797
+ "grad_norm": 13796.53125,
798
+ "learning_rate": 8.848508814233106e-05,
799
+ "loss": 0.9476,
800
+ "step": 30500
801
+ },
802
+ {
803
+ "epoch": 13.449023861171366,
804
+ "grad_norm": 5713.0361328125,
805
+ "learning_rate": 8.812080266430921e-05,
806
+ "loss": 0.9417,
807
+ "step": 31000
808
+ },
809
+ {
810
+ "epoch": 13.66594360086768,
811
+ "grad_norm": 3508.00634765625,
812
+ "learning_rate": 8.77508745092958e-05,
813
+ "loss": 0.9446,
814
+ "step": 31500
815
+ },
816
+ {
817
+ "epoch": 13.882863340563992,
818
+ "grad_norm": 4413.595703125,
819
+ "learning_rate": 8.737607648009131e-05,
820
+ "loss": 0.9488,
821
+ "step": 32000
822
+ },
823
+ {
824
+ "epoch": 14.0,
825
+ "eval_loss": 0.9241105318069458,
826
+ "eval_map": 0.4602,
827
+ "eval_map_50": 0.8064,
828
+ "eval_map_75": 0.4696,
829
+ "eval_map_bus": 0.5309,
830
+ "eval_map_car": 0.4761,
831
+ "eval_map_container": 0.5062,
832
+ "eval_map_large": 0.6817,
833
+ "eval_map_medium": 0.4503,
834
+ "eval_map_motorbike": 0.3275,
835
+ "eval_map_small": 0.1122,
836
+ "eval_mar_1": 0.2462,
837
+ "eval_mar_10": 0.5333,
838
+ "eval_mar_100": 0.5819,
839
+ "eval_mar_100_bus": 0.6633,
840
+ "eval_mar_100_car": 0.5828,
841
+ "eval_mar_100_container": 0.6259,
842
+ "eval_mar_100_motorbike": 0.4557,
843
+ "eval_mar_large": 0.7655,
844
+ "eval_mar_medium": 0.5804,
845
+ "eval_mar_small": 0.2896,
846
+ "eval_runtime": 145.7229,
847
+ "eval_samples_per_second": 15.804,
848
+ "eval_steps_per_second": 1.976,
849
+ "step": 32270
850
+ },
851
+ {
852
+ "epoch": 14.099783080260304,
853
+ "grad_norm": 7677.90478515625,
854
+ "learning_rate": 8.699645692575454e-05,
855
+ "loss": 0.9519,
856
+ "step": 32500
857
+ },
858
+ {
859
+ "epoch": 14.316702819956616,
860
+ "grad_norm": 4366.18212890625,
861
+ "learning_rate": 8.661283833169138e-05,
862
+ "loss": 0.9471,
863
+ "step": 33000
864
+ },
865
+ {
866
+ "epoch": 14.533622559652928,
867
+ "grad_norm": 5411.373046875,
868
+ "learning_rate": 8.622373265190294e-05,
869
+ "loss": 0.9423,
870
+ "step": 33500
871
+ },
872
+ {
873
+ "epoch": 14.750542299349242,
874
+ "grad_norm": 7791.236328125,
875
+ "learning_rate": 8.582995409968335e-05,
876
+ "loss": 0.9357,
877
+ "step": 34000
878
+ },
879
+ {
880
+ "epoch": 14.967462039045554,
881
+ "grad_norm": 20878.41796875,
882
+ "learning_rate": 8.543155347258457e-05,
883
+ "loss": 0.9371,
884
+ "step": 34500
885
+ },
886
+ {
887
+ "epoch": 15.0,
888
+ "eval_loss": 0.9158653020858765,
889
+ "eval_map": 0.4646,
890
+ "eval_map_50": 0.8109,
891
+ "eval_map_75": 0.479,
892
+ "eval_map_bus": 0.525,
893
+ "eval_map_car": 0.4942,
894
+ "eval_map_container": 0.5118,
895
+ "eval_map_large": 0.6799,
896
+ "eval_map_medium": 0.4567,
897
+ "eval_map_motorbike": 0.3276,
898
+ "eval_map_small": 0.1141,
899
+ "eval_mar_1": 0.2484,
900
+ "eval_mar_10": 0.533,
901
+ "eval_mar_100": 0.5795,
902
+ "eval_mar_100_bus": 0.6523,
903
+ "eval_mar_100_car": 0.5909,
904
+ "eval_mar_100_container": 0.6278,
905
+ "eval_mar_100_motorbike": 0.447,
906
+ "eval_mar_large": 0.7745,
907
+ "eval_mar_medium": 0.5767,
908
+ "eval_mar_small": 0.3179,
909
+ "eval_runtime": 143.8949,
910
+ "eval_samples_per_second": 16.005,
911
+ "eval_steps_per_second": 2.001,
912
+ "step": 34575
913
+ },
914
+ {
915
+ "epoch": 15.184381778741866,
916
+ "grad_norm": 8804.869140625,
917
+ "learning_rate": 8.502858216440758e-05,
918
+ "loss": 0.9249,
919
+ "step": 35000
920
+ },
921
+ {
922
+ "epoch": 15.401301518438178,
923
+ "grad_norm": 7941.79443359375,
924
+ "learning_rate": 8.462191161340177e-05,
925
+ "loss": 0.9341,
926
+ "step": 35500
927
+ },
928
+ {
929
+ "epoch": 15.61822125813449,
930
+ "grad_norm": 11279.1513671875,
931
+ "learning_rate": 8.420996435565979e-05,
932
+ "loss": 0.9342,
933
+ "step": 36000
934
+ },
935
+ {
936
+ "epoch": 15.835140997830802,
937
+ "grad_norm": 2519.469970703125,
938
+ "learning_rate": 8.379360400220391e-05,
939
+ "loss": 0.9311,
940
+ "step": 36500
941
+ },
942
+ {
943
+ "epoch": 16.0,
944
+ "eval_loss": 0.8999801278114319,
945
+ "eval_map": 0.4741,
946
+ "eval_map_50": 0.8169,
947
+ "eval_map_75": 0.4893,
948
+ "eval_map_bus": 0.5392,
949
+ "eval_map_car": 0.5006,
950
+ "eval_map_container": 0.5198,
951
+ "eval_map_large": 0.6914,
952
+ "eval_map_medium": 0.4673,
953
+ "eval_map_motorbike": 0.3368,
954
+ "eval_map_small": 0.1224,
955
+ "eval_mar_1": 0.2523,
956
+ "eval_mar_10": 0.544,
957
+ "eval_mar_100": 0.5909,
958
+ "eval_mar_100_bus": 0.6792,
959
+ "eval_mar_100_car": 0.5917,
960
+ "eval_mar_100_container": 0.6356,
961
+ "eval_mar_100_motorbike": 0.4571,
962
+ "eval_mar_large": 0.7817,
963
+ "eval_mar_medium": 0.593,
964
+ "eval_mar_small": 0.3192,
965
+ "eval_runtime": 144.4917,
966
+ "eval_samples_per_second": 15.939,
967
+ "eval_steps_per_second": 1.993,
968
+ "step": 36880
969
+ },
970
+ {
971
+ "epoch": 16.052060737527114,
972
+ "grad_norm": 6870.65673828125,
973
+ "learning_rate": 8.337288426364522e-05,
974
+ "loss": 0.9394,
975
+ "step": 37000
976
+ },
977
+ {
978
+ "epoch": 16.268980477223426,
979
+ "grad_norm": 18748.70703125,
980
+ "learning_rate": 8.294785941295685e-05,
981
+ "loss": 0.9206,
982
+ "step": 37500
983
+ },
984
+ {
985
+ "epoch": 16.485900216919738,
986
+ "grad_norm": 8770.451171875,
987
+ "learning_rate": 8.251944703380834e-05,
988
+ "loss": 0.9277,
989
+ "step": 38000
990
+ },
991
+ {
992
+ "epoch": 16.702819956616054,
993
+ "grad_norm": 10527.71484375,
994
+ "learning_rate": 8.208598532633761e-05,
995
+ "loss": 0.9363,
996
+ "step": 38500
997
+ },
998
+ {
999
+ "epoch": 16.919739696312366,
1000
+ "grad_norm": 13457.677734375,
1001
+ "learning_rate": 8.164838451708801e-05,
1002
+ "loss": 0.9284,
1003
+ "step": 39000
1004
+ },
1005
+ {
1006
+ "epoch": 17.0,
1007
+ "eval_loss": 0.9076481461524963,
1008
+ "eval_map": 0.472,
1009
+ "eval_map_50": 0.8148,
1010
+ "eval_map_75": 0.4845,
1011
+ "eval_map_bus": 0.5397,
1012
+ "eval_map_car": 0.4969,
1013
+ "eval_map_container": 0.5162,
1014
+ "eval_map_large": 0.684,
1015
+ "eval_map_medium": 0.4625,
1016
+ "eval_map_motorbike": 0.3354,
1017
+ "eval_map_small": 0.1174,
1018
+ "eval_mar_1": 0.252,
1019
+ "eval_mar_10": 0.5415,
1020
+ "eval_mar_100": 0.5905,
1021
+ "eval_mar_100_bus": 0.6732,
1022
+ "eval_mar_100_car": 0.5938,
1023
+ "eval_mar_100_container": 0.6354,
1024
+ "eval_mar_100_motorbike": 0.4596,
1025
+ "eval_mar_large": 0.767,
1026
+ "eval_mar_medium": 0.5905,
1027
+ "eval_mar_small": 0.315,
1028
+ "eval_runtime": 145.0152,
1029
+ "eval_samples_per_second": 15.881,
1030
+ "eval_steps_per_second": 1.986,
1031
+ "step": 39185
1032
+ },
1033
+ {
1034
+ "epoch": 17.136659436008678,
1035
+ "grad_norm": 6826.60595703125,
1036
+ "learning_rate": 8.120670105669583e-05,
1037
+ "loss": 0.9237,
1038
+ "step": 39500
1039
+ },
1040
+ {
1041
+ "epoch": 17.35357917570499,
1042
+ "grad_norm": 10753.5400390625,
1043
+ "learning_rate": 8.076099192246061e-05,
1044
+ "loss": 0.9257,
1045
+ "step": 40000
1046
+ },
1047
+ {
1048
+ "epoch": 17.570498915401302,
1049
+ "grad_norm": 5758.92919921875,
1050
+ "learning_rate": 8.031131461099501e-05,
1051
+ "loss": 0.9247,
1052
+ "step": 40500
1053
+ },
1054
+ {
1055
+ "epoch": 17.787418655097614,
1056
+ "grad_norm": 11779.173828125,
1057
+ "learning_rate": 7.985772713080777e-05,
1058
+ "loss": 0.9107,
1059
+ "step": 41000
1060
+ },
1061
+ {
1062
+ "epoch": 18.0,
1063
+ "eval_loss": 0.9001684188842773,
1064
+ "eval_map": 0.4753,
1065
+ "eval_map_50": 0.8198,
1066
+ "eval_map_75": 0.4952,
1067
+ "eval_map_bus": 0.5451,
1068
+ "eval_map_car": 0.501,
1069
+ "eval_map_container": 0.5198,
1070
+ "eval_map_large": 0.6944,
1071
+ "eval_map_medium": 0.4679,
1072
+ "eval_map_motorbike": 0.3352,
1073
+ "eval_map_small": 0.1207,
1074
+ "eval_mar_1": 0.2516,
1075
+ "eval_mar_10": 0.5428,
1076
+ "eval_mar_100": 0.5902,
1077
+ "eval_mar_100_bus": 0.6716,
1078
+ "eval_mar_100_car": 0.5977,
1079
+ "eval_mar_100_container": 0.6353,
1080
+ "eval_mar_100_motorbike": 0.4562,
1081
+ "eval_mar_large": 0.7743,
1082
+ "eval_mar_medium": 0.592,
1083
+ "eval_mar_small": 0.3133,
1084
+ "eval_runtime": 143.7248,
1085
+ "eval_samples_per_second": 16.024,
1086
+ "eval_steps_per_second": 2.004,
1087
+ "step": 41490
1088
+ },
1089
+ {
1090
+ "epoch": 18.004338394793926,
1091
+ "grad_norm": 3542.92626953125,
1092
+ "learning_rate": 7.940120667790506e-05,
1093
+ "loss": 0.919,
1094
+ "step": 41500
1095
+ },
1096
+ {
1097
+ "epoch": 18.22125813449024,
1098
+ "grad_norm": 22165.880859375,
1099
+ "learning_rate": 7.893998242197806e-05,
1100
+ "loss": 0.928,
1101
+ "step": 42000
1102
+ },
1103
+ {
1104
+ "epoch": 18.43817787418655,
1105
+ "grad_norm": 3429.292236328125,
1106
+ "learning_rate": 7.84750248995952e-05,
1107
+ "loss": 0.9192,
1108
+ "step": 42500
1109
+ },
1110
+ {
1111
+ "epoch": 18.655097613882862,
1112
+ "grad_norm": 10154.87109375,
1113
+ "learning_rate": 7.800639409041694e-05,
1114
+ "loss": 0.9183,
1115
+ "step": 43000
1116
+ },
1117
+ {
1118
+ "epoch": 18.872017353579174,
1119
+ "grad_norm": 25265.55078125,
1120
+ "learning_rate": 7.753415044795898e-05,
1121
+ "loss": 0.9125,
1122
+ "step": 43500
1123
+ },
1124
+ {
1125
+ "epoch": 19.0,
1126
+ "eval_loss": 0.8902826905250549,
1127
+ "eval_map": 0.4805,
1128
+ "eval_map_50": 0.8223,
1129
+ "eval_map_75": 0.5006,
1130
+ "eval_map_bus": 0.5537,
1131
+ "eval_map_car": 0.5047,
1132
+ "eval_map_container": 0.5198,
1133
+ "eval_map_large": 0.6876,
1134
+ "eval_map_medium": 0.4736,
1135
+ "eval_map_motorbike": 0.3438,
1136
+ "eval_map_small": 0.119,
1137
+ "eval_mar_1": 0.2528,
1138
+ "eval_mar_10": 0.5449,
1139
+ "eval_mar_100": 0.5921,
1140
+ "eval_mar_100_bus": 0.6688,
1141
+ "eval_mar_100_car": 0.6008,
1142
+ "eval_mar_100_container": 0.6319,
1143
+ "eval_mar_100_motorbike": 0.4669,
1144
+ "eval_mar_large": 0.7716,
1145
+ "eval_mar_medium": 0.5905,
1146
+ "eval_mar_small": 0.3164,
1147
+ "eval_runtime": 146.715,
1148
+ "eval_samples_per_second": 15.697,
1149
+ "eval_steps_per_second": 1.963,
1150
+ "step": 43795
1151
+ },
1152
+ {
1153
+ "epoch": 19.088937093275486,
1154
+ "grad_norm": 7347.31640625,
1155
+ "learning_rate": 7.705835489179352e-05,
1156
+ "loss": 0.9033,
1157
+ "step": 44000
1158
+ },
1159
+ {
1160
+ "epoch": 19.305856832971802,
1161
+ "grad_norm": 5815.42822265625,
1162
+ "learning_rate": 7.658003081440886e-05,
1163
+ "loss": 0.9049,
1164
+ "step": 44500
1165
+ },
1166
+ {
1167
+ "epoch": 19.522776572668114,
1168
+ "grad_norm": 5560.4921875,
1169
+ "learning_rate": 7.609732280983515e-05,
1170
+ "loss": 0.9124,
1171
+ "step": 45000
1172
+ },
1173
+ {
1174
+ "epoch": 19.739696312364426,
1175
+ "grad_norm": 5130.99462890625,
1176
+ "learning_rate": 7.56112482427524e-05,
1177
+ "loss": 0.9205,
1178
+ "step": 45500
1179
+ },
1180
+ {
1181
+ "epoch": 19.95661605206074,
1182
+ "grad_norm": 17505.642578125,
1183
+ "learning_rate": 7.512186981692624e-05,
1184
+ "loss": 0.913,
1185
+ "step": 46000
1186
+ },
1187
+ {
1188
+ "epoch": 20.0,
1189
+ "eval_loss": 0.8936046957969666,
1190
+ "eval_map": 0.4822,
1191
+ "eval_map_50": 0.8243,
1192
+ "eval_map_75": 0.504,
1193
+ "eval_map_bus": 0.5592,
1194
+ "eval_map_car": 0.4984,
1195
+ "eval_map_container": 0.5236,
1196
+ "eval_map_large": 0.6904,
1197
+ "eval_map_medium": 0.4777,
1198
+ "eval_map_motorbike": 0.3475,
1199
+ "eval_map_small": 0.115,
1200
+ "eval_mar_1": 0.2536,
1201
+ "eval_mar_10": 0.5453,
1202
+ "eval_mar_100": 0.5923,
1203
+ "eval_mar_100_bus": 0.6727,
1204
+ "eval_mar_100_car": 0.5953,
1205
+ "eval_mar_100_container": 0.6347,
1206
+ "eval_mar_100_motorbike": 0.4666,
1207
+ "eval_mar_large": 0.7683,
1208
+ "eval_mar_medium": 0.5946,
1209
+ "eval_mar_small": 0.3124,
1210
+ "eval_runtime": 146.1172,
1211
+ "eval_samples_per_second": 15.761,
1212
+ "eval_steps_per_second": 1.971,
1213
+ "step": 46100
1214
+ },
1215
+ {
1216
+ "epoch": 20.17353579175705,
1217
+ "grad_norm": 21176.666015625,
1218
+ "learning_rate": 7.462925066232113e-05,
1219
+ "loss": 0.9039,
1220
+ "step": 46500
1221
+ },
1222
+ {
1223
+ "epoch": 20.390455531453362,
1224
+ "grad_norm": 4085.164306640625,
1225
+ "learning_rate": 7.413444904801033e-05,
1226
+ "loss": 0.9082,
1227
+ "step": 47000
1228
+ },
1229
+ {
1230
+ "epoch": 20.607375271149674,
1231
+ "grad_norm": 20771.896484375,
1232
+ "learning_rate": 7.363554565211042e-05,
1233
+ "loss": 0.9087,
1234
+ "step": 47500
1235
+ },
1236
+ {
1237
+ "epoch": 20.824295010845987,
1238
+ "grad_norm": 5990.69580078125,
1239
+ "learning_rate": 7.313359326369684e-05,
1240
+ "loss": 0.908,
1241
+ "step": 48000
1242
+ },
1243
+ {
1244
+ "epoch": 21.0,
1245
+ "eval_loss": 0.8810420036315918,
1246
+ "eval_map": 0.483,
1247
+ "eval_map_50": 0.8249,
1248
+ "eval_map_75": 0.5047,
1249
+ "eval_map_bus": 0.5531,
1250
+ "eval_map_car": 0.5081,
1251
+ "eval_map_container": 0.5229,
1252
+ "eval_map_large": 0.7091,
1253
+ "eval_map_medium": 0.4743,
1254
+ "eval_map_motorbike": 0.348,
1255
+ "eval_map_small": 0.1233,
1256
+ "eval_mar_1": 0.2539,
1257
+ "eval_mar_10": 0.5486,
1258
+ "eval_mar_100": 0.5987,
1259
+ "eval_mar_100_bus": 0.6764,
1260
+ "eval_mar_100_car": 0.6079,
1261
+ "eval_mar_100_container": 0.6426,
1262
+ "eval_mar_100_motorbike": 0.4679,
1263
+ "eval_mar_large": 0.7947,
1264
+ "eval_mar_medium": 0.5992,
1265
+ "eval_mar_small": 0.3117,
1266
+ "eval_runtime": 143.798,
1267
+ "eval_samples_per_second": 16.016,
1268
+ "eval_steps_per_second": 2.003,
1269
+ "step": 48405
1270
+ },
1271
+ {
1272
+ "epoch": 21.0412147505423,
1273
+ "grad_norm": 4679.26220703125,
1274
+ "learning_rate": 7.262865663477902e-05,
1275
+ "loss": 0.9023,
1276
+ "step": 48500
1277
+ },
1278
+ {
1279
+ "epoch": 21.25813449023861,
1280
+ "grad_norm": 18654.177734375,
1281
+ "learning_rate": 7.21208009023343e-05,
1282
+ "loss": 0.9069,
1283
+ "step": 49000
1284
+ },
1285
+ {
1286
+ "epoch": 21.475054229934923,
1287
+ "grad_norm": 10939.3623046875,
1288
+ "learning_rate": 7.161009157990521e-05,
1289
+ "loss": 0.8956,
1290
+ "step": 49500
1291
+ },
1292
+ {
1293
+ "epoch": 21.691973969631235,
1294
+ "grad_norm": 14381.9765625,
1295
+ "learning_rate": 7.109659454914833e-05,
1296
+ "loss": 0.8925,
1297
+ "step": 50000
1298
+ },
1299
+ {
1300
+ "epoch": 21.90889370932755,
1301
+ "grad_norm": 5005.86865234375,
1302
+ "learning_rate": 7.058141116015095e-05,
1303
+ "loss": 0.8997,
1304
+ "step": 50500
1305
+ },
1306
+ {
1307
+ "epoch": 22.0,
1308
+ "eval_loss": 0.8783358335494995,
1309
+ "eval_map": 0.4846,
1310
+ "eval_map_50": 0.8275,
1311
+ "eval_map_75": 0.5024,
1312
+ "eval_map_bus": 0.5507,
1313
+ "eval_map_car": 0.5086,
1314
+ "eval_map_container": 0.5299,
1315
+ "eval_map_large": 0.699,
1316
+ "eval_map_medium": 0.4761,
1317
+ "eval_map_motorbike": 0.3491,
1318
+ "eval_map_small": 0.1256,
1319
+ "eval_mar_1": 0.2542,
1320
+ "eval_mar_10": 0.5503,
1321
+ "eval_mar_100": 0.5994,
1322
+ "eval_mar_100_bus": 0.6731,
1323
+ "eval_mar_100_car": 0.6068,
1324
+ "eval_mar_100_container": 0.6486,
1325
+ "eval_mar_100_motorbike": 0.469,
1326
+ "eval_mar_large": 0.7854,
1327
+ "eval_mar_medium": 0.5993,
1328
+ "eval_mar_small": 0.3314,
1329
+ "eval_runtime": 144.6256,
1330
+ "eval_samples_per_second": 15.924,
1331
+ "eval_steps_per_second": 1.991,
1332
+ "step": 50710
1333
+ },
1334
+ {
1335
+ "epoch": 22.125813449023862,
1336
+ "grad_norm": 19755.140625,
1337
+ "learning_rate": 7.006254303068497e-05,
1338
+ "loss": 0.8975,
1339
+ "step": 51000
1340
+ },
1341
+ {
1342
+ "epoch": 22.342733188720175,
1343
+ "grad_norm": 11582.6494140625,
1344
+ "learning_rate": 6.954108682712056e-05,
1345
+ "loss": 0.898,
1346
+ "step": 51500
1347
+ },
1348
+ {
1349
+ "epoch": 22.559652928416487,
1350
+ "grad_norm": 6819.56396484375,
1351
+ "learning_rate": 6.901710981746509e-05,
1352
+ "loss": 0.8999,
1353
+ "step": 52000
1354
+ },
1355
+ {
1356
+ "epoch": 22.7765726681128,
1357
+ "grad_norm": 7609.8115234375,
1358
+ "learning_rate": 6.849067959491079e-05,
1359
+ "loss": 0.8978,
1360
+ "step": 52500
1361
+ },
1362
+ {
1363
+ "epoch": 22.99349240780911,
1364
+ "grad_norm": 10856.244140625,
1365
+ "learning_rate": 6.796186406911505e-05,
1366
+ "loss": 0.8865,
1367
+ "step": 53000
1368
+ },
1369
+ {
1370
+ "epoch": 23.0,
1371
+ "eval_loss": 0.871816098690033,
1372
+ "eval_map": 0.4901,
1373
+ "eval_map_50": 0.8289,
1374
+ "eval_map_75": 0.5202,
1375
+ "eval_map_bus": 0.5591,
1376
+ "eval_map_car": 0.5153,
1377
+ "eval_map_container": 0.5306,
1378
+ "eval_map_large": 0.7037,
1379
+ "eval_map_medium": 0.4841,
1380
+ "eval_map_motorbike": 0.3553,
1381
+ "eval_map_small": 0.1243,
1382
+ "eval_mar_1": 0.2558,
1383
+ "eval_mar_10": 0.5526,
1384
+ "eval_mar_100": 0.6008,
1385
+ "eval_mar_100_bus": 0.6757,
1386
+ "eval_mar_100_car": 0.6077,
1387
+ "eval_mar_100_container": 0.6452,
1388
+ "eval_mar_100_motorbike": 0.4746,
1389
+ "eval_mar_large": 0.788,
1390
+ "eval_mar_medium": 0.6019,
1391
+ "eval_mar_small": 0.3201,
1392
+ "eval_runtime": 145.8422,
1393
+ "eval_samples_per_second": 15.791,
1394
+ "eval_steps_per_second": 1.975,
1395
+ "step": 53015
1396
+ },
1397
+ {
1398
+ "epoch": 23.210412147505423,
1399
+ "grad_norm": 6373.787109375,
1400
+ "learning_rate": 6.743286051281191e-05,
1401
+ "loss": 0.8974,
1402
+ "step": 53500
1403
+ },
1404
+ {
1405
+ "epoch": 23.427331887201735,
1406
+ "grad_norm": 3711.38671875,
1407
+ "learning_rate": 6.68994881889314e-05,
1408
+ "loss": 0.9048,
1409
+ "step": 54000
1410
+ },
1411
+ {
1412
+ "epoch": 23.644251626898047,
1413
+ "grad_norm": 1683.364501953125,
1414
+ "learning_rate": 6.636393582598046e-05,
1415
+ "loss": 0.911,
1416
+ "step": 54500
1417
+ },
1418
+ {
1419
+ "epoch": 23.86117136659436,
1420
+ "grad_norm": 2381.84423828125,
1421
+ "learning_rate": 6.582627251037533e-05,
1422
+ "loss": 0.9052,
1423
+ "step": 55000
1424
+ },
1425
+ {
1426
+ "epoch": 24.0,
1427
+ "eval_loss": 0.8951303958892822,
1428
+ "eval_map": 0.4779,
1429
+ "eval_map_50": 0.8254,
1430
+ "eval_map_75": 0.4914,
1431
+ "eval_map_bus": 0.5509,
1432
+ "eval_map_car": 0.5002,
1433
+ "eval_map_container": 0.5227,
1434
+ "eval_map_large": 0.697,
1435
+ "eval_map_medium": 0.469,
1436
+ "eval_map_motorbike": 0.3376,
1437
+ "eval_map_small": 0.1198,
1438
+ "eval_mar_1": 0.2524,
1439
+ "eval_mar_10": 0.5426,
1440
+ "eval_mar_100": 0.5902,
1441
+ "eval_mar_100_bus": 0.6715,
1442
+ "eval_mar_100_car": 0.5914,
1443
+ "eval_mar_100_container": 0.643,
1444
+ "eval_mar_100_motorbike": 0.455,
1445
+ "eval_mar_large": 0.7785,
1446
+ "eval_mar_medium": 0.5885,
1447
+ "eval_mar_small": 0.319,
1448
+ "eval_runtime": 145.8551,
1449
+ "eval_samples_per_second": 15.79,
1450
+ "eval_steps_per_second": 1.975,
1451
+ "step": 55320
1452
+ },
1453
+ {
1454
+ "epoch": 24.07809110629067,
1455
+ "grad_norm": 19113.990234375,
1456
+ "learning_rate": 6.528656760084578e-05,
1457
+ "loss": 0.9112,
1458
+ "step": 55500
1459
+ },
1460
+ {
1461
+ "epoch": 24.295010845986983,
1462
+ "grad_norm": 20675.185546875,
1463
+ "learning_rate": 6.474489071948787e-05,
1464
+ "loss": 0.8944,
1465
+ "step": 56000
1466
+ },
1467
+ {
1468
+ "epoch": 24.5119305856833,
1469
+ "grad_norm": 10454.0859375,
1470
+ "learning_rate": 6.420131174278254e-05,
1471
+ "loss": 0.8915,
1472
+ "step": 56500
1473
+ },
1474
+ {
1475
+ "epoch": 24.72885032537961,
1476
+ "grad_norm": 4770.7431640625,
1477
+ "learning_rate": 6.365590079258165e-05,
1478
+ "loss": 0.8891,
1479
+ "step": 57000
1480
+ },
1481
+ {
1482
+ "epoch": 24.945770065075923,
1483
+ "grad_norm": 16598.36328125,
1484
+ "learning_rate": 6.310872822706217e-05,
1485
+ "loss": 0.8982,
1486
+ "step": 57500
1487
+ },
1488
+ {
1489
+ "epoch": 25.0,
1490
+ "eval_loss": 0.8864476084709167,
1491
+ "eval_map": 0.4662,
1492
+ "eval_map_50": 0.8105,
1493
+ "eval_map_75": 0.4884,
1494
+ "eval_map_bus": 0.515,
1495
+ "eval_map_car": 0.4889,
1496
+ "eval_map_container": 0.5084,
1497
+ "eval_map_large": 0.6737,
1498
+ "eval_map_medium": 0.4711,
1499
+ "eval_map_motorbike": 0.3524,
1500
+ "eval_map_small": 0.1286,
1501
+ "eval_mar_1": 0.2482,
1502
+ "eval_mar_10": 0.5444,
1503
+ "eval_mar_100": 0.5934,
1504
+ "eval_mar_100_bus": 0.6696,
1505
+ "eval_mar_100_car": 0.5993,
1506
+ "eval_mar_100_container": 0.6371,
1507
+ "eval_mar_100_motorbike": 0.4675,
1508
+ "eval_mar_large": 0.7729,
1509
+ "eval_mar_medium": 0.5969,
1510
+ "eval_mar_small": 0.3226,
1511
+ "eval_runtime": 144.7854,
1512
+ "eval_samples_per_second": 15.906,
1513
+ "eval_steps_per_second": 1.989,
1514
+ "step": 57625
1515
+ },
1516
+ {
1517
+ "epoch": 25.162689804772235,
1518
+ "grad_norm": 6422.17041015625,
1519
+ "learning_rate": 6.25609639994632e-05,
1520
+ "loss": 0.8784,
1521
+ "step": 58000
1522
+ },
1523
+ {
1524
+ "epoch": 25.379609544468547,
1525
+ "grad_norm": 16689.212890625,
1526
+ "learning_rate": 6.201048334737784e-05,
1527
+ "loss": 0.8855,
1528
+ "step": 58500
1529
+ },
1530
+ {
1531
+ "epoch": 25.59652928416486,
1532
+ "grad_norm": 18894.322265625,
1533
+ "learning_rate": 6.145845333932029e-05,
1534
+ "loss": 0.8872,
1535
+ "step": 59000
1536
+ },
1537
+ {
1538
+ "epoch": 25.81344902386117,
1539
+ "grad_norm": 4359.09521484375,
1540
+ "learning_rate": 6.090494518732801e-05,
1541
+ "loss": 0.8835,
1542
+ "step": 59500
1543
+ },
1544
+ {
1545
+ "epoch": 26.0,
1546
+ "eval_loss": 0.8596058487892151,
1547
+ "eval_map": 0.4972,
1548
+ "eval_map_50": 0.8376,
1549
+ "eval_map_75": 0.5257,
1550
+ "eval_map_bus": 0.5714,
1551
+ "eval_map_car": 0.5186,
1552
+ "eval_map_container": 0.5391,
1553
+ "eval_map_large": 0.7028,
1554
+ "eval_map_medium": 0.4915,
1555
+ "eval_map_motorbike": 0.3598,
1556
+ "eval_map_small": 0.1342,
1557
+ "eval_mar_1": 0.2583,
1558
+ "eval_mar_10": 0.5574,
1559
+ "eval_mar_100": 0.6057,
1560
+ "eval_mar_100_bus": 0.6883,
1561
+ "eval_mar_100_car": 0.6094,
1562
+ "eval_mar_100_container": 0.6514,
1563
+ "eval_mar_100_motorbike": 0.4737,
1564
+ "eval_mar_large": 0.7866,
1565
+ "eval_mar_medium": 0.6078,
1566
+ "eval_mar_small": 0.3419,
1567
+ "eval_runtime": 145.0724,
1568
+ "eval_samples_per_second": 15.875,
1569
+ "eval_steps_per_second": 1.985,
1570
+ "step": 59930
1571
+ },
1572
+ {
1573
+ "epoch": 26.030368763557483,
1574
+ "grad_norm": 28658.181640625,
1575
+ "learning_rate": 6.0350030294119464e-05,
1576
+ "loss": 0.8825,
1577
+ "step": 60000
1578
+ },
1579
+ {
1580
+ "epoch": 26.247288503253795,
1581
+ "grad_norm": 6283.97509765625,
1582
+ "learning_rate": 5.979489402880387e-05,
1583
+ "loss": 0.8833,
1584
+ "step": 60500
1585
+ },
1586
+ {
1587
+ "epoch": 26.464208242950107,
1588
+ "grad_norm": 4272.13037109375,
1589
+ "learning_rate": 5.923738303304234e-05,
1590
+ "loss": 0.8844,
1591
+ "step": 61000
1592
+ },
1593
+ {
1594
+ "epoch": 26.68112798264642,
1595
+ "grad_norm": 14047.6572265625,
1596
+ "learning_rate": 5.8678680412085076e-05,
1597
+ "loss": 0.8785,
1598
+ "step": 61500
1599
+ },
1600
+ {
1601
+ "epoch": 26.89804772234273,
1602
+ "grad_norm": 8514.8544921875,
1603
+ "learning_rate": 5.811885823873863e-05,
1604
+ "loss": 0.8692,
1605
+ "step": 62000
1606
+ },
1607
+ {
1608
+ "epoch": 27.0,
1609
+ "eval_loss": 0.8606675863265991,
1610
+ "eval_map": 0.4927,
1611
+ "eval_map_50": 0.837,
1612
+ "eval_map_75": 0.5186,
1613
+ "eval_map_bus": 0.5641,
1614
+ "eval_map_car": 0.5122,
1615
+ "eval_map_container": 0.5369,
1616
+ "eval_map_large": 0.7052,
1617
+ "eval_map_medium": 0.4874,
1618
+ "eval_map_motorbike": 0.3576,
1619
+ "eval_map_small": 0.1426,
1620
+ "eval_mar_1": 0.2578,
1621
+ "eval_mar_10": 0.5566,
1622
+ "eval_mar_100": 0.6043,
1623
+ "eval_mar_100_bus": 0.6885,
1624
+ "eval_mar_100_car": 0.6043,
1625
+ "eval_mar_100_container": 0.6507,
1626
+ "eval_mar_100_motorbike": 0.4737,
1627
+ "eval_mar_large": 0.7936,
1628
+ "eval_mar_medium": 0.608,
1629
+ "eval_mar_small": 0.3251,
1630
+ "eval_runtime": 145.9029,
1631
+ "eval_samples_per_second": 15.784,
1632
+ "eval_steps_per_second": 1.974,
1633
+ "step": 62235
1634
+ },
1635
+ {
1636
+ "epoch": 27.114967462039047,
1637
+ "grad_norm": 17592.478515625,
1638
+ "learning_rate": 5.755798873023209e-05,
1639
+ "loss": 0.8719,
1640
+ "step": 62500
1641
+ },
1642
+ {
1643
+ "epoch": 27.33188720173536,
1644
+ "grad_norm": 28705.041015625,
1645
+ "learning_rate": 5.699614423890112e-05,
1646
+ "loss": 0.873,
1647
+ "step": 63000
1648
+ },
1649
+ {
1650
+ "epoch": 27.54880694143167,
1651
+ "grad_norm": 14412.8916015625,
1652
+ "learning_rate": 5.643339724285447e-05,
1653
+ "loss": 0.8654,
1654
+ "step": 63500
1655
+ },
1656
+ {
1657
+ "epoch": 27.765726681127983,
1658
+ "grad_norm": 11298.9375,
1659
+ "learning_rate": 5.587094827038153e-05,
1660
+ "loss": 0.8829,
1661
+ "step": 64000
1662
+ },
1663
+ {
1664
+ "epoch": 27.982646420824295,
1665
+ "grad_norm": 8732.7822265625,
1666
+ "learning_rate": 5.530661559735304e-05,
1667
+ "loss": 0.8694,
1668
+ "step": 64500
1669
+ },
1670
+ {
1671
+ "epoch": 28.0,
1672
+ "eval_loss": 0.8637914657592773,
1673
+ "eval_map": 0.4929,
1674
+ "eval_map_50": 0.8355,
1675
+ "eval_map_75": 0.5204,
1676
+ "eval_map_bus": 0.5621,
1677
+ "eval_map_car": 0.5152,
1678
+ "eval_map_container": 0.5343,
1679
+ "eval_map_large": 0.7119,
1680
+ "eval_map_medium": 0.4885,
1681
+ "eval_map_motorbike": 0.3601,
1682
+ "eval_map_small": 0.1308,
1683
+ "eval_mar_1": 0.2571,
1684
+ "eval_mar_10": 0.556,
1685
+ "eval_mar_100": 0.6029,
1686
+ "eval_mar_100_bus": 0.6851,
1687
+ "eval_mar_100_car": 0.6085,
1688
+ "eval_mar_100_container": 0.6431,
1689
+ "eval_mar_100_motorbike": 0.4749,
1690
+ "eval_mar_large": 0.7964,
1691
+ "eval_mar_medium": 0.6042,
1692
+ "eval_mar_small": 0.3162,
1693
+ "eval_runtime": 145.47,
1694
+ "eval_samples_per_second": 15.831,
1695
+ "eval_steps_per_second": 1.98,
1696
+ "step": 64540
1697
+ },
1698
+ {
1699
+ "epoch": 28.199566160520607,
1700
+ "grad_norm": 6033.21923828125,
1701
+ "learning_rate": 5.474159836931288e-05,
1702
+ "loss": 0.8805,
1703
+ "step": 65000
1704
+ },
1705
+ {
1706
+ "epoch": 28.41648590021692,
1707
+ "grad_norm": 7858.236328125,
1708
+ "learning_rate": 5.417596947365383e-05,
1709
+ "loss": 0.878,
1710
+ "step": 65500
1711
+ },
1712
+ {
1713
+ "epoch": 28.63340563991323,
1714
+ "grad_norm": 11765.7236328125,
1715
+ "learning_rate": 5.361093470095784e-05,
1716
+ "loss": 0.8746,
1717
+ "step": 66000
1718
+ },
1719
+ {
1720
+ "epoch": 28.850325379609544,
1721
+ "grad_norm": 3453.400146484375,
1722
+ "learning_rate": 5.304430229684755e-05,
1723
+ "loss": 0.8744,
1724
+ "step": 66500
1725
+ },
1726
+ {
1727
+ "epoch": 29.0,
1728
+ "eval_loss": 0.8675053119659424,
1729
+ "eval_map": 0.4883,
1730
+ "eval_map_50": 0.8358,
1731
+ "eval_map_75": 0.5122,
1732
+ "eval_map_bus": 0.5564,
1733
+ "eval_map_car": 0.5131,
1734
+ "eval_map_container": 0.5244,
1735
+ "eval_map_large": 0.6955,
1736
+ "eval_map_medium": 0.4818,
1737
+ "eval_map_motorbike": 0.3594,
1738
+ "eval_map_small": 0.1305,
1739
+ "eval_mar_1": 0.2542,
1740
+ "eval_mar_10": 0.5483,
1741
+ "eval_mar_100": 0.5948,
1742
+ "eval_mar_100_bus": 0.6693,
1743
+ "eval_mar_100_car": 0.602,
1744
+ "eval_mar_100_container": 0.6352,
1745
+ "eval_mar_100_motorbike": 0.4727,
1746
+ "eval_mar_large": 0.7802,
1747
+ "eval_mar_medium": 0.5936,
1748
+ "eval_mar_small": 0.3363,
1749
+ "eval_runtime": 145.2093,
1750
+ "eval_samples_per_second": 15.86,
1751
+ "eval_steps_per_second": 1.983,
1752
+ "step": 66845
1753
+ },
1754
+ {
1755
+ "epoch": 29.067245119305856,
1756
+ "grad_norm": 4124.80615234375,
1757
+ "learning_rate": 5.24772771768233e-05,
1758
+ "loss": 0.8763,
1759
+ "step": 67000
1760
+ },
1761
+ {
1762
+ "epoch": 29.284164859002168,
1763
+ "grad_norm": 12881.8603515625,
1764
+ "learning_rate": 5.1909932487296506e-05,
1765
+ "loss": 0.8783,
1766
+ "step": 67500
1767
+ },
1768
+ {
1769
+ "epoch": 29.501084598698483,
1770
+ "grad_norm": 11561.8046875,
1771
+ "learning_rate": 5.134234141590315e-05,
1772
+ "loss": 0.867,
1773
+ "step": 68000
1774
+ },
1775
+ {
1776
+ "epoch": 29.718004338394795,
1777
+ "grad_norm": 4195.20458984375,
1778
+ "learning_rate": 5.077571283466763e-05,
1779
+ "loss": 0.8623,
1780
+ "step": 68500
1781
+ },
1782
+ {
1783
+ "epoch": 29.934924078091107,
1784
+ "grad_norm": 6831.716796875,
1785
+ "learning_rate": 5.020784880687298e-05,
1786
+ "loss": 0.8608,
1787
+ "step": 69000
1788
+ },
1789
+ {
1790
+ "epoch": 30.0,
1791
+ "eval_loss": 0.8547199964523315,
1792
+ "eval_map": 0.5,
1793
+ "eval_map_50": 0.8401,
1794
+ "eval_map_75": 0.5304,
1795
+ "eval_map_bus": 0.5762,
1796
+ "eval_map_car": 0.5197,
1797
+ "eval_map_container": 0.5424,
1798
+ "eval_map_large": 0.7118,
1799
+ "eval_map_medium": 0.4962,
1800
+ "eval_map_motorbike": 0.3617,
1801
+ "eval_map_small": 0.1291,
1802
+ "eval_mar_1": 0.2596,
1803
+ "eval_mar_10": 0.559,
1804
+ "eval_mar_100": 0.6079,
1805
+ "eval_mar_100_bus": 0.6877,
1806
+ "eval_mar_100_car": 0.6139,
1807
+ "eval_mar_100_container": 0.6542,
1808
+ "eval_mar_100_motorbike": 0.4756,
1809
+ "eval_mar_large": 0.7898,
1810
+ "eval_mar_medium": 0.6093,
1811
+ "eval_mar_small": 0.34,
1812
+ "eval_runtime": 144.8308,
1813
+ "eval_samples_per_second": 15.901,
1814
+ "eval_steps_per_second": 1.989,
1815
+ "step": 69150
1816
+ },
1817
+ {
1818
+ "epoch": 30.15184381778742,
1819
+ "grad_norm": 9134.2763671875,
1820
+ "learning_rate": 4.9639957966519604e-05,
1821
+ "loss": 0.8683,
1822
+ "step": 69500
1823
+ },
1824
+ {
1825
+ "epoch": 30.36876355748373,
1826
+ "grad_norm": 6325.7216796875,
1827
+ "learning_rate": 4.907211357169706e-05,
1828
+ "loss": 0.8653,
1829
+ "step": 70000
1830
+ },
1831
+ {
1832
+ "epoch": 30.585683297180044,
1833
+ "grad_norm": 13625.853515625,
1834
+ "learning_rate": 4.850438887450347e-05,
1835
+ "loss": 0.8714,
1836
+ "step": 70500
1837
+ },
1838
+ {
1839
+ "epoch": 30.802603036876356,
1840
+ "grad_norm": 15019.505859375,
1841
+ "learning_rate": 4.793799193390669e-05,
1842
+ "loss": 0.8566,
1843
+ "step": 71000
1844
+ },
1845
+ {
1846
+ "epoch": 31.0,
1847
+ "eval_loss": 0.8513031005859375,
1848
+ "eval_map": 0.5,
1849
+ "eval_map_50": 0.8418,
1850
+ "eval_map_75": 0.5333,
1851
+ "eval_map_bus": 0.5714,
1852
+ "eval_map_car": 0.5239,
1853
+ "eval_map_container": 0.5422,
1854
+ "eval_map_large": 0.7031,
1855
+ "eval_map_medium": 0.4972,
1856
+ "eval_map_motorbike": 0.3626,
1857
+ "eval_map_small": 0.1282,
1858
+ "eval_mar_1": 0.2582,
1859
+ "eval_mar_10": 0.5594,
1860
+ "eval_mar_100": 0.6082,
1861
+ "eval_mar_100_bus": 0.6856,
1862
+ "eval_mar_100_car": 0.6153,
1863
+ "eval_mar_100_container": 0.6545,
1864
+ "eval_mar_100_motorbike": 0.4772,
1865
+ "eval_mar_large": 0.7875,
1866
+ "eval_mar_medium": 0.6116,
1867
+ "eval_mar_small": 0.3505,
1868
+ "eval_runtime": 144.5939,
1869
+ "eval_samples_per_second": 15.927,
1870
+ "eval_steps_per_second": 1.992,
1871
+ "step": 71455
1872
+ },
1873
+ {
1874
+ "epoch": 31.019522776572668,
1875
+ "grad_norm": 9826.4052734375,
1876
+ "learning_rate": 4.7370725711717407e-05,
1877
+ "loss": 0.8693,
1878
+ "step": 71500
1879
+ },
1880
+ {
1881
+ "epoch": 31.23644251626898,
1882
+ "grad_norm": 3330.38623046875,
1883
+ "learning_rate": 4.6803798666703856e-05,
1884
+ "loss": 0.8625,
1885
+ "step": 72000
1886
+ },
1887
+ {
1888
+ "epoch": 31.453362255965292,
1889
+ "grad_norm": 4638.60107421875,
1890
+ "learning_rate": 4.623728393262576e-05,
1891
+ "loss": 0.8672,
1892
+ "step": 72500
1893
+ },
1894
+ {
1895
+ "epoch": 31.670281995661604,
1896
+ "grad_norm": 10907.923828125,
1897
+ "learning_rate": 4.5672386115774904e-05,
1898
+ "loss": 0.8625,
1899
+ "step": 73000
1900
+ },
1901
+ {
1902
+ "epoch": 31.887201735357916,
1903
+ "grad_norm": 3609.915283203125,
1904
+ "learning_rate": 4.5106913993021883e-05,
1905
+ "loss": 0.8586,
1906
+ "step": 73500
1907
+ },
1908
+ {
1909
+ "epoch": 32.0,
1910
+ "eval_loss": 0.8532082438468933,
1911
+ "eval_map": 0.5035,
1912
+ "eval_map_50": 0.8414,
1913
+ "eval_map_75": 0.538,
1914
+ "eval_map_bus": 0.5799,
1915
+ "eval_map_car": 0.5235,
1916
+ "eval_map_container": 0.5481,
1917
+ "eval_map_large": 0.7084,
1918
+ "eval_map_medium": 0.4993,
1919
+ "eval_map_motorbike": 0.3626,
1920
+ "eval_map_small": 0.1292,
1921
+ "eval_mar_1": 0.2603,
1922
+ "eval_mar_10": 0.5621,
1923
+ "eval_mar_100": 0.6094,
1924
+ "eval_mar_100_bus": 0.6893,
1925
+ "eval_mar_100_car": 0.6139,
1926
+ "eval_mar_100_container": 0.6587,
1927
+ "eval_mar_100_motorbike": 0.4758,
1928
+ "eval_mar_large": 0.7924,
1929
+ "eval_mar_medium": 0.6135,
1930
+ "eval_mar_small": 0.3286,
1931
+ "eval_runtime": 145.9881,
1932
+ "eval_samples_per_second": 15.775,
1933
+ "eval_steps_per_second": 1.973,
1934
+ "step": 73760
1935
+ },
1936
+ {
1937
+ "epoch": 32.10412147505423,
1938
+ "grad_norm": 11651.8828125,
1939
+ "learning_rate": 4.4542073079838684e-05,
1940
+ "loss": 0.8618,
1941
+ "step": 74000
1942
+ },
1943
+ {
1944
+ "epoch": 32.321041214750544,
1945
+ "grad_norm": 17436.388671875,
1946
+ "learning_rate": 4.39779362408734e-05,
1947
+ "loss": 0.8542,
1948
+ "step": 74500
1949
+ },
1950
+ {
1951
+ "epoch": 32.53796095444685,
1952
+ "grad_norm": 20459.701171875,
1953
+ "learning_rate": 4.341457624994828e-05,
1954
+ "loss": 0.8554,
1955
+ "step": 75000
1956
+ },
1957
+ {
1958
+ "epoch": 32.75488069414317,
1959
+ "grad_norm": 5393.4375,
1960
+ "learning_rate": 4.285318990553778e-05,
1961
+ "loss": 0.8528,
1962
+ "step": 75500
1963
+ },
1964
+ {
1965
+ "epoch": 32.971800433839476,
1966
+ "grad_norm": 4986.16796875,
1967
+ "learning_rate": 4.229159960541767e-05,
1968
+ "loss": 0.8598,
1969
+ "step": 76000
1970
+ },
1971
+ {
1972
+ "epoch": 33.0,
1973
+ "eval_loss": 0.8493744134902954,
1974
+ "eval_map": 0.499,
1975
+ "eval_map_50": 0.8415,
1976
+ "eval_map_75": 0.5323,
1977
+ "eval_map_bus": 0.5708,
1978
+ "eval_map_car": 0.5215,
1979
+ "eval_map_container": 0.5432,
1980
+ "eval_map_large": 0.7104,
1981
+ "eval_map_medium": 0.4969,
1982
+ "eval_map_motorbike": 0.3606,
1983
+ "eval_map_small": 0.1301,
1984
+ "eval_mar_1": 0.2582,
1985
+ "eval_mar_10": 0.5605,
1986
+ "eval_mar_100": 0.6084,
1987
+ "eval_mar_100_bus": 0.6857,
1988
+ "eval_mar_100_car": 0.615,
1989
+ "eval_mar_100_container": 0.6551,
1990
+ "eval_mar_100_motorbike": 0.4778,
1991
+ "eval_mar_large": 0.7942,
1992
+ "eval_mar_medium": 0.6113,
1993
+ "eval_mar_small": 0.3277,
1994
+ "eval_runtime": 144.7755,
1995
+ "eval_samples_per_second": 15.907,
1996
+ "eval_steps_per_second": 1.989,
1997
+ "step": 76065
1998
+ },
1999
+ {
2000
+ "epoch": 33.18872017353579,
2001
+ "grad_norm": 6646.677734375,
2002
+ "learning_rate": 4.173212384261402e-05,
2003
+ "loss": 0.8671,
2004
+ "step": 76500
2005
+ },
2006
+ {
2007
+ "epoch": 33.40563991323211,
2008
+ "grad_norm": 4728.1552734375,
2009
+ "learning_rate": 4.1172592425996225e-05,
2010
+ "loss": 0.8671,
2011
+ "step": 77000
2012
+ },
2013
+ {
2014
+ "epoch": 33.622559652928416,
2015
+ "grad_norm": 9825.3720703125,
2016
+ "learning_rate": 4.0614199747616364e-05,
2017
+ "loss": 0.8668,
2018
+ "step": 77500
2019
+ },
2020
+ {
2021
+ "epoch": 33.83947939262473,
2022
+ "grad_norm": 3324.343505859375,
2023
+ "learning_rate": 4.0057017840298235e-05,
2024
+ "loss": 0.8728,
2025
+ "step": 78000
2026
+ },
2027
+ {
2028
+ "epoch": 34.0,
2029
+ "eval_loss": 0.8495970368385315,
2030
+ "eval_map": 0.5027,
2031
+ "eval_map_50": 0.841,
2032
+ "eval_map_75": 0.5357,
2033
+ "eval_map_bus": 0.5809,
2034
+ "eval_map_car": 0.5196,
2035
+ "eval_map_container": 0.5452,
2036
+ "eval_map_large": 0.715,
2037
+ "eval_map_medium": 0.4969,
2038
+ "eval_map_motorbike": 0.3652,
2039
+ "eval_map_small": 0.1332,
2040
+ "eval_mar_1": 0.2609,
2041
+ "eval_mar_10": 0.5617,
2042
+ "eval_mar_100": 0.6105,
2043
+ "eval_mar_100_bus": 0.6941,
2044
+ "eval_mar_100_car": 0.6116,
2045
+ "eval_mar_100_container": 0.6565,
2046
+ "eval_mar_100_motorbike": 0.4799,
2047
+ "eval_mar_large": 0.7994,
2048
+ "eval_mar_medium": 0.6147,
2049
+ "eval_mar_small": 0.3335,
2050
+ "eval_runtime": 145.9053,
2051
+ "eval_samples_per_second": 15.784,
2052
+ "eval_steps_per_second": 1.974,
2053
+ "step": 78370
2054
+ }
2055
+ ],
2056
+ "logging_steps": 500,
2057
+ "max_steps": 138300,
2058
+ "num_input_tokens_seen": 0,
2059
+ "num_train_epochs": 60,
2060
+ "save_steps": 500,
2061
+ "stateful_callbacks": {
2062
+ "TrainerControl": {
2063
+ "args": {
2064
+ "should_epoch_stop": false,
2065
+ "should_evaluate": false,
2066
+ "should_log": false,
2067
+ "should_save": true,
2068
+ "should_training_stop": false
2069
+ },
2070
+ "attributes": {}
2071
+ }
2072
+ },
2073
+ "total_flos": 1.4412665205261022e+20,
2074
+ "train_batch_size": 4,
2075
+ "trial_name": null,
2076
+ "trial_params": null
2077
+ }
checkpoint-78370/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e79ce30220cc0f492cce991ef1b82af549d14310ccde1c3ab2fa0987f62e3936
3
+ size 5240
checkpoint-80675/config.json ADDED
@@ -0,0 +1,67 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "datamonster/ms-cond-detr-res-50-vehicles",
3
+ "activation_dropout": 0.0,
4
+ "activation_function": "relu",
5
+ "architectures": [
6
+ "ConditionalDetrForObjectDetection"
7
+ ],
8
+ "attention_dropout": 0.0,
9
+ "auxiliary_loss": false,
10
+ "backbone": "resnet50",
11
+ "backbone_config": null,
12
+ "backbone_kwargs": {
13
+ "in_chans": 3,
14
+ "out_indices": [
15
+ 1,
16
+ 2,
17
+ 3,
18
+ 4
19
+ ]
20
+ },
21
+ "bbox_cost": 5,
22
+ "bbox_loss_coefficient": 5,
23
+ "class_cost": 2,
24
+ "cls_loss_coefficient": 2,
25
+ "d_model": 256,
26
+ "decoder_attention_heads": 8,
27
+ "decoder_ffn_dim": 2048,
28
+ "decoder_layerdrop": 0.0,
29
+ "decoder_layers": 6,
30
+ "dice_loss_coefficient": 1,
31
+ "dilation": false,
32
+ "dropout": 0.1,
33
+ "encoder_attention_heads": 8,
34
+ "encoder_ffn_dim": 2048,
35
+ "encoder_layerdrop": 0.0,
36
+ "encoder_layers": 6,
37
+ "focal_alpha": 0.25,
38
+ "giou_cost": 2,
39
+ "giou_loss_coefficient": 2,
40
+ "id2label": {
41
+ "0": "motorbike",
42
+ "1": "car",
43
+ "2": "bus",
44
+ "3": "container"
45
+ },
46
+ "init_std": 0.02,
47
+ "init_xavier_std": 1.0,
48
+ "is_encoder_decoder": true,
49
+ "label2id": {
50
+ "bus": 2,
51
+ "car": 1,
52
+ "container": 3,
53
+ "motorbike": 0
54
+ },
55
+ "mask_loss_coefficient": 1,
56
+ "max_position_embeddings": 1024,
57
+ "model_type": "conditional_detr",
58
+ "num_channels": 3,
59
+ "num_hidden_layers": 6,
60
+ "num_queries": 300,
61
+ "position_embedding_type": "sine",
62
+ "scale_embedding": false,
63
+ "torch_dtype": "float32",
64
+ "transformers_version": "4.45.1",
65
+ "use_pretrained_backbone": true,
66
+ "use_timm_backbone": true
67
+ }
checkpoint-80675/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:da7170a2c2b873313751cb57af69aba3b8bf53bd675edc48aef4d7d6d4924934
3
+ size 174078768
checkpoint-80675/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fa370e08b460fbbe63cb6f41e649e94ac3062c4a828427a65511a259d6197915
3
+ size 187563450
checkpoint-80675/preprocessor_config.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_convert_annotations": true,
3
+ "do_normalize": true,
4
+ "do_pad": true,
5
+ "do_rescale": true,
6
+ "do_resize": true,
7
+ "format": "coco_detection",
8
+ "image_mean": [
9
+ 0.485,
10
+ 0.456,
11
+ 0.406
12
+ ],
13
+ "image_processor_type": "ConditionalDetrImageProcessor",
14
+ "image_std": [
15
+ 0.229,
16
+ 0.224,
17
+ 0.225
18
+ ],
19
+ "pad_size": {
20
+ "height": 768,
21
+ "width": 768
22
+ },
23
+ "resample": 2,
24
+ "rescale_factor": 0.00392156862745098,
25
+ "size": {
26
+ "max_height": 768,
27
+ "max_width": 768
28
+ }
29
+ }
checkpoint-80675/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c64a6d4b6b11cd39db34f2ce4c221e060d52a7d8336691f5b42a0892c217978d
3
+ size 14244
checkpoint-80675/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a8ae33b3f44e469dd54aef44e110b9fddab8566f63f13ba5bcf03098b35ceaa7
3
+ size 1000
checkpoint-80675/trainer_state.json ADDED
@@ -0,0 +1,2140 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.5081,
3
+ "best_model_checkpoint": "ms-cond-detr-res-50-vehicles/checkpoint-80675",
4
+ "epoch": 35.0,
5
+ "eval_steps": 500,
6
+ "global_step": 80675,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.21691973969631237,
13
+ "grad_norm": 38753.67578125,
14
+ "learning_rate": 9.999686465963656e-05,
15
+ "loss": 1.2341,
16
+ "step": 500
17
+ },
18
+ {
19
+ "epoch": 0.43383947939262474,
20
+ "grad_norm": 14156.765625,
21
+ "learning_rate": 9.998728034076959e-05,
22
+ "loss": 1.1441,
23
+ "step": 1000
24
+ },
25
+ {
26
+ "epoch": 0.6507592190889371,
27
+ "grad_norm": 7688.85986328125,
28
+ "learning_rate": 9.99712861472983e-05,
29
+ "loss": 1.1357,
30
+ "step": 1500
31
+ },
32
+ {
33
+ "epoch": 0.8676789587852495,
34
+ "grad_norm": 10963.5341796875,
35
+ "learning_rate": 9.994882003818517e-05,
36
+ "loss": 1.1322,
37
+ "step": 2000
38
+ },
39
+ {
40
+ "epoch": 1.0,
41
+ "eval_loss": 1.0705492496490479,
42
+ "eval_map": 0.3886,
43
+ "eval_map_50": 0.7276,
44
+ "eval_map_75": 0.371,
45
+ "eval_map_bus": 0.4553,
46
+ "eval_map_car": 0.4267,
47
+ "eval_map_container": 0.4051,
48
+ "eval_map_large": 0.6092,
49
+ "eval_map_medium": 0.3725,
50
+ "eval_map_motorbike": 0.2674,
51
+ "eval_map_small": 0.0755,
52
+ "eval_mar_1": 0.2228,
53
+ "eval_mar_10": 0.4736,
54
+ "eval_mar_100": 0.5226,
55
+ "eval_mar_100_bus": 0.6014,
56
+ "eval_mar_100_car": 0.5398,
57
+ "eval_mar_100_container": 0.5541,
58
+ "eval_mar_100_motorbike": 0.395,
59
+ "eval_mar_large": 0.7004,
60
+ "eval_mar_medium": 0.5176,
61
+ "eval_mar_small": 0.2475,
62
+ "eval_runtime": 149.8641,
63
+ "eval_samples_per_second": 15.367,
64
+ "eval_steps_per_second": 1.922,
65
+ "step": 2305
66
+ },
67
+ {
68
+ "epoch": 1.0845986984815619,
69
+ "grad_norm": 8778.8525390625,
70
+ "learning_rate": 9.991991051623087e-05,
71
+ "loss": 1.1204,
72
+ "step": 2500
73
+ },
74
+ {
75
+ "epoch": 1.3015184381778742,
76
+ "grad_norm": 8215.1748046875,
77
+ "learning_rate": 9.988456131077249e-05,
78
+ "loss": 1.0925,
79
+ "step": 3000
80
+ },
81
+ {
82
+ "epoch": 1.5184381778741867,
83
+ "grad_norm": 15722.8212890625,
84
+ "learning_rate": 9.984277698186814e-05,
85
+ "loss": 1.0867,
86
+ "step": 3500
87
+ },
88
+ {
89
+ "epoch": 1.735357917570499,
90
+ "grad_norm": 17670.681640625,
91
+ "learning_rate": 9.979456291970887e-05,
92
+ "loss": 1.0714,
93
+ "step": 4000
94
+ },
95
+ {
96
+ "epoch": 1.9522776572668112,
97
+ "grad_norm": 15503.669921875,
98
+ "learning_rate": 9.974004102519452e-05,
99
+ "loss": 1.0605,
100
+ "step": 4500
101
+ },
102
+ {
103
+ "epoch": 2.0,
104
+ "eval_loss": 1.0418221950531006,
105
+ "eval_map": 0.3977,
106
+ "eval_map_50": 0.7371,
107
+ "eval_map_75": 0.389,
108
+ "eval_map_bus": 0.4529,
109
+ "eval_map_car": 0.4318,
110
+ "eval_map_container": 0.4316,
111
+ "eval_map_large": 0.6212,
112
+ "eval_map_medium": 0.3842,
113
+ "eval_map_motorbike": 0.2745,
114
+ "eval_map_small": 0.0841,
115
+ "eval_mar_1": 0.2265,
116
+ "eval_mar_10": 0.4923,
117
+ "eval_mar_100": 0.5393,
118
+ "eval_mar_100_bus": 0.6241,
119
+ "eval_mar_100_car": 0.5529,
120
+ "eval_mar_100_container": 0.5848,
121
+ "eval_mar_100_motorbike": 0.3954,
122
+ "eval_mar_large": 0.7101,
123
+ "eval_mar_medium": 0.5389,
124
+ "eval_mar_small": 0.2511,
125
+ "eval_runtime": 146.5587,
126
+ "eval_samples_per_second": 15.714,
127
+ "eval_steps_per_second": 1.965,
128
+ "step": 4610
129
+ },
130
+ {
131
+ "epoch": 2.1691973969631237,
132
+ "grad_norm": 3109.66162109375,
133
+ "learning_rate": 9.967899980939251e-05,
134
+ "loss": 1.0591,
135
+ "step": 5000
136
+ },
137
+ {
138
+ "epoch": 2.3861171366594363,
139
+ "grad_norm": 5812.068359375,
140
+ "learning_rate": 9.961154998764017e-05,
141
+ "loss": 1.0676,
142
+ "step": 5500
143
+ },
144
+ {
145
+ "epoch": 2.6030368763557483,
146
+ "grad_norm": 7711.87158203125,
147
+ "learning_rate": 9.953770026098485e-05,
148
+ "loss": 1.0673,
149
+ "step": 6000
150
+ },
151
+ {
152
+ "epoch": 2.819956616052061,
153
+ "grad_norm": 11271.8916015625,
154
+ "learning_rate": 9.945746015606358e-05,
155
+ "loss": 1.0634,
156
+ "step": 6500
157
+ },
158
+ {
159
+ "epoch": 3.0,
160
+ "eval_loss": 1.0088645219802856,
161
+ "eval_map": 0.4231,
162
+ "eval_map_50": 0.758,
163
+ "eval_map_75": 0.4223,
164
+ "eval_map_bus": 0.4788,
165
+ "eval_map_car": 0.4565,
166
+ "eval_map_container": 0.4648,
167
+ "eval_map_large": 0.658,
168
+ "eval_map_medium": 0.4084,
169
+ "eval_map_motorbike": 0.2923,
170
+ "eval_map_small": 0.09,
171
+ "eval_mar_1": 0.2352,
172
+ "eval_mar_10": 0.5048,
173
+ "eval_mar_100": 0.555,
174
+ "eval_mar_100_bus": 0.6323,
175
+ "eval_mar_100_car": 0.5672,
176
+ "eval_mar_100_container": 0.6028,
177
+ "eval_mar_100_motorbike": 0.4178,
178
+ "eval_mar_large": 0.7546,
179
+ "eval_mar_medium": 0.5521,
180
+ "eval_mar_small": 0.2802,
181
+ "eval_runtime": 147.4693,
182
+ "eval_samples_per_second": 15.617,
183
+ "eval_steps_per_second": 1.953,
184
+ "step": 6915
185
+ },
186
+ {
187
+ "epoch": 3.036876355748373,
188
+ "grad_norm": 20180.71484375,
189
+ "learning_rate": 9.937084002387407e-05,
190
+ "loss": 1.0418,
191
+ "step": 7000
192
+ },
193
+ {
194
+ "epoch": 3.2537960954446854,
195
+ "grad_norm": 11818.376953125,
196
+ "learning_rate": 9.927804336468942e-05,
197
+ "loss": 1.0386,
198
+ "step": 7500
199
+ },
200
+ {
201
+ "epoch": 3.470715835140998,
202
+ "grad_norm": 9070.5185546875,
203
+ "learning_rate": 9.917871022281522e-05,
204
+ "loss": 1.0635,
205
+ "step": 8000
206
+ },
207
+ {
208
+ "epoch": 3.6876355748373104,
209
+ "grad_norm": 7446.32666015625,
210
+ "learning_rate": 9.907303301249819e-05,
211
+ "loss": 1.0466,
212
+ "step": 8500
213
+ },
214
+ {
215
+ "epoch": 3.9045553145336225,
216
+ "grad_norm": 2428.818603515625,
217
+ "learning_rate": 9.896102536613032e-05,
218
+ "loss": 1.0288,
219
+ "step": 9000
220
+ },
221
+ {
222
+ "epoch": 4.0,
223
+ "eval_loss": 1.002821922302246,
224
+ "eval_map": 0.4297,
225
+ "eval_map_50": 0.7647,
226
+ "eval_map_75": 0.429,
227
+ "eval_map_bus": 0.4925,
228
+ "eval_map_car": 0.4582,
229
+ "eval_map_container": 0.4685,
230
+ "eval_map_large": 0.6667,
231
+ "eval_map_medium": 0.4184,
232
+ "eval_map_motorbike": 0.2997,
233
+ "eval_map_small": 0.0816,
234
+ "eval_mar_1": 0.2383,
235
+ "eval_mar_10": 0.5069,
236
+ "eval_mar_100": 0.5549,
237
+ "eval_mar_100_bus": 0.634,
238
+ "eval_mar_100_car": 0.5649,
239
+ "eval_mar_100_container": 0.5965,
240
+ "eval_mar_100_motorbike": 0.4243,
241
+ "eval_mar_large": 0.7579,
242
+ "eval_mar_medium": 0.5516,
243
+ "eval_mar_small": 0.2596,
244
+ "eval_runtime": 146.3143,
245
+ "eval_samples_per_second": 15.74,
246
+ "eval_steps_per_second": 1.968,
247
+ "step": 9220
248
+ },
249
+ {
250
+ "epoch": 4.1214750542299345,
251
+ "grad_norm": 4431.853515625,
252
+ "learning_rate": 9.88429446733437e-05,
253
+ "loss": 1.0289,
254
+ "step": 9500
255
+ },
256
+ {
257
+ "epoch": 4.3383947939262475,
258
+ "grad_norm": 16948.7890625,
259
+ "learning_rate": 9.871833290237144e-05,
260
+ "loss": 1.031,
261
+ "step": 10000
262
+ },
263
+ {
264
+ "epoch": 4.55531453362256,
265
+ "grad_norm": 5605.24951171875,
266
+ "learning_rate": 9.858743645176947e-05,
267
+ "loss": 1.0303,
268
+ "step": 10500
269
+ },
270
+ {
271
+ "epoch": 4.7722342733188725,
272
+ "grad_norm": 16235.091796875,
273
+ "learning_rate": 9.845027220721935e-05,
274
+ "loss": 1.0206,
275
+ "step": 11000
276
+ },
277
+ {
278
+ "epoch": 4.989154013015185,
279
+ "grad_norm": 20350.36328125,
280
+ "learning_rate": 9.830685786294999e-05,
281
+ "loss": 1.0226,
282
+ "step": 11500
283
+ },
284
+ {
285
+ "epoch": 5.0,
286
+ "eval_loss": 0.9774508476257324,
287
+ "eval_map": 0.4366,
288
+ "eval_map_50": 0.7785,
289
+ "eval_map_75": 0.4388,
290
+ "eval_map_bus": 0.5023,
291
+ "eval_map_car": 0.4658,
292
+ "eval_map_container": 0.4766,
293
+ "eval_map_large": 0.6605,
294
+ "eval_map_medium": 0.4243,
295
+ "eval_map_motorbike": 0.3015,
296
+ "eval_map_small": 0.0942,
297
+ "eval_mar_1": 0.2402,
298
+ "eval_mar_10": 0.5173,
299
+ "eval_mar_100": 0.5666,
300
+ "eval_mar_100_bus": 0.6534,
301
+ "eval_mar_100_car": 0.5747,
302
+ "eval_mar_100_container": 0.6145,
303
+ "eval_mar_100_motorbike": 0.4239,
304
+ "eval_mar_large": 0.751,
305
+ "eval_mar_medium": 0.5659,
306
+ "eval_mar_small": 0.2757,
307
+ "eval_runtime": 146.1171,
308
+ "eval_samples_per_second": 15.761,
309
+ "eval_steps_per_second": 1.971,
310
+ "step": 11525
311
+ },
312
+ {
313
+ "epoch": 5.206073752711497,
314
+ "grad_norm": 11680.388671875,
315
+ "learning_rate": 9.815721191945502e-05,
316
+ "loss": 1.0021,
317
+ "step": 12000
318
+ },
319
+ {
320
+ "epoch": 5.422993492407809,
321
+ "grad_norm": 6468.74072265625,
322
+ "learning_rate": 9.800167158422275e-05,
323
+ "loss": 1.0049,
324
+ "step": 12500
325
+ },
326
+ {
327
+ "epoch": 5.639913232104122,
328
+ "grad_norm": 10755.380859375,
329
+ "learning_rate": 9.78396335205612e-05,
330
+ "loss": 1.016,
331
+ "step": 13000
332
+ },
333
+ {
334
+ "epoch": 5.856832971800434,
335
+ "grad_norm": 6588.42578125,
336
+ "learning_rate": 9.767142412975505e-05,
337
+ "loss": 1.022,
338
+ "step": 13500
339
+ },
340
+ {
341
+ "epoch": 6.0,
342
+ "eval_loss": 0.9850358366966248,
343
+ "eval_map": 0.4356,
344
+ "eval_map_50": 0.7775,
345
+ "eval_map_75": 0.4345,
346
+ "eval_map_bus": 0.5032,
347
+ "eval_map_car": 0.455,
348
+ "eval_map_container": 0.4783,
349
+ "eval_map_large": 0.661,
350
+ "eval_map_medium": 0.4213,
351
+ "eval_map_motorbike": 0.3059,
352
+ "eval_map_small": 0.106,
353
+ "eval_mar_1": 0.2402,
354
+ "eval_mar_10": 0.514,
355
+ "eval_mar_100": 0.563,
356
+ "eval_mar_100_bus": 0.651,
357
+ "eval_mar_100_car": 0.5596,
358
+ "eval_mar_100_container": 0.6163,
359
+ "eval_mar_100_motorbike": 0.4251,
360
+ "eval_mar_large": 0.7479,
361
+ "eval_mar_medium": 0.5634,
362
+ "eval_mar_small": 0.2541,
363
+ "eval_runtime": 145.7736,
364
+ "eval_samples_per_second": 15.798,
365
+ "eval_steps_per_second": 1.976,
366
+ "step": 13830
367
+ },
368
+ {
369
+ "epoch": 6.073752711496746,
370
+ "grad_norm": 8365.4345703125,
371
+ "learning_rate": 9.749706511086642e-05,
372
+ "loss": 1.0065,
373
+ "step": 14000
374
+ },
375
+ {
376
+ "epoch": 6.290672451193059,
377
+ "grad_norm": 7092.89306640625,
378
+ "learning_rate": 9.731694602810796e-05,
379
+ "loss": 1.0004,
380
+ "step": 14500
381
+ },
382
+ {
383
+ "epoch": 6.507592190889371,
384
+ "grad_norm": 7779.22314453125,
385
+ "learning_rate": 9.713036820450053e-05,
386
+ "loss": 1.0193,
387
+ "step": 15000
388
+ },
389
+ {
390
+ "epoch": 6.724511930585683,
391
+ "grad_norm": 5161.82958984375,
392
+ "learning_rate": 9.693771054918806e-05,
393
+ "loss": 0.9992,
394
+ "step": 15500
395
+ },
396
+ {
397
+ "epoch": 6.941431670281996,
398
+ "grad_norm": 5342.78662109375,
399
+ "learning_rate": 9.673899791506617e-05,
400
+ "loss": 0.9936,
401
+ "step": 16000
402
+ },
403
+ {
404
+ "epoch": 7.0,
405
+ "eval_loss": 0.9677829742431641,
406
+ "eval_map": 0.4394,
407
+ "eval_map_50": 0.781,
408
+ "eval_map_75": 0.4415,
409
+ "eval_map_bus": 0.5059,
410
+ "eval_map_car": 0.4716,
411
+ "eval_map_container": 0.4731,
412
+ "eval_map_large": 0.6636,
413
+ "eval_map_medium": 0.4308,
414
+ "eval_map_motorbike": 0.307,
415
+ "eval_map_small": 0.109,
416
+ "eval_mar_1": 0.2401,
417
+ "eval_mar_10": 0.5184,
418
+ "eval_mar_100": 0.5674,
419
+ "eval_mar_100_bus": 0.6546,
420
+ "eval_mar_100_car": 0.5736,
421
+ "eval_mar_100_container": 0.6126,
422
+ "eval_mar_100_motorbike": 0.4289,
423
+ "eval_mar_large": 0.7468,
424
+ "eval_mar_medium": 0.5696,
425
+ "eval_mar_small": 0.2775,
426
+ "eval_runtime": 147.1057,
427
+ "eval_samples_per_second": 15.655,
428
+ "eval_steps_per_second": 1.958,
429
+ "step": 16135
430
+ },
431
+ {
432
+ "epoch": 7.158351409978308,
433
+ "grad_norm": 17945.982421875,
434
+ "learning_rate": 9.653425593612456e-05,
435
+ "loss": 0.9817,
436
+ "step": 16500
437
+ },
438
+ {
439
+ "epoch": 7.375271149674621,
440
+ "grad_norm": 2631.81591796875,
441
+ "learning_rate": 9.632393848695017e-05,
442
+ "loss": 0.9845,
443
+ "step": 17000
444
+ },
445
+ {
446
+ "epoch": 7.592190889370933,
447
+ "grad_norm": 1825.7994384765625,
448
+ "learning_rate": 9.61072297519293e-05,
449
+ "loss": 0.9839,
450
+ "step": 17500
451
+ },
452
+ {
453
+ "epoch": 7.809110629067245,
454
+ "grad_norm": 8929.837890625,
455
+ "learning_rate": 9.588457317037219e-05,
456
+ "loss": 0.9908,
457
+ "step": 18000
458
+ },
459
+ {
460
+ "epoch": 8.0,
461
+ "eval_loss": 0.9544239640235901,
462
+ "eval_map": 0.4515,
463
+ "eval_map_50": 0.7929,
464
+ "eval_map_75": 0.459,
465
+ "eval_map_bus": 0.5215,
466
+ "eval_map_car": 0.4759,
467
+ "eval_map_container": 0.4956,
468
+ "eval_map_large": 0.6707,
469
+ "eval_map_medium": 0.4407,
470
+ "eval_map_motorbike": 0.3131,
471
+ "eval_map_small": 0.103,
472
+ "eval_mar_1": 0.2439,
473
+ "eval_mar_10": 0.5243,
474
+ "eval_mar_100": 0.573,
475
+ "eval_mar_100_bus": 0.6558,
476
+ "eval_mar_100_car": 0.579,
477
+ "eval_mar_100_container": 0.6217,
478
+ "eval_mar_100_motorbike": 0.4355,
479
+ "eval_mar_large": 0.7575,
480
+ "eval_mar_medium": 0.5708,
481
+ "eval_mar_small": 0.3107,
482
+ "eval_runtime": 147.1314,
483
+ "eval_samples_per_second": 15.653,
484
+ "eval_steps_per_second": 1.957,
485
+ "step": 18440
486
+ },
487
+ {
488
+ "epoch": 8.026030368763557,
489
+ "grad_norm": 9851.580078125,
490
+ "learning_rate": 9.565599746504495e-05,
491
+ "loss": 0.9914,
492
+ "step": 18500
493
+ },
494
+ {
495
+ "epoch": 8.242950108459869,
496
+ "grad_norm": 46131.0703125,
497
+ "learning_rate": 9.542153212228251e-05,
498
+ "loss": 0.9789,
499
+ "step": 19000
500
+ },
501
+ {
502
+ "epoch": 8.459869848156183,
503
+ "grad_norm": 40477.8671875,
504
+ "learning_rate": 9.518120738818473e-05,
505
+ "loss": 0.9642,
506
+ "step": 19500
507
+ },
508
+ {
509
+ "epoch": 8.676789587852495,
510
+ "grad_norm": 21795.466796875,
511
+ "learning_rate": 9.493505426471479e-05,
512
+ "loss": 0.9637,
513
+ "step": 20000
514
+ },
515
+ {
516
+ "epoch": 8.893709327548807,
517
+ "grad_norm": 5294.40673828125,
518
+ "learning_rate": 9.468361416878137e-05,
519
+ "loss": 0.9853,
520
+ "step": 20500
521
+ },
522
+ {
523
+ "epoch": 9.0,
524
+ "eval_loss": 0.9467804431915283,
525
+ "eval_map": 0.4529,
526
+ "eval_map_50": 0.7951,
527
+ "eval_map_75": 0.46,
528
+ "eval_map_bus": 0.5273,
529
+ "eval_map_car": 0.4728,
530
+ "eval_map_container": 0.4969,
531
+ "eval_map_large": 0.6735,
532
+ "eval_map_medium": 0.4421,
533
+ "eval_map_motorbike": 0.3144,
534
+ "eval_map_small": 0.1016,
535
+ "eval_mar_1": 0.2466,
536
+ "eval_mar_10": 0.5271,
537
+ "eval_mar_100": 0.5772,
538
+ "eval_mar_100_bus": 0.6592,
539
+ "eval_mar_100_car": 0.5831,
540
+ "eval_mar_100_container": 0.6281,
541
+ "eval_mar_100_motorbike": 0.4385,
542
+ "eval_mar_large": 0.7607,
543
+ "eval_mar_medium": 0.5767,
544
+ "eval_mar_small": 0.2947,
545
+ "eval_runtime": 146.7224,
546
+ "eval_samples_per_second": 15.696,
547
+ "eval_steps_per_second": 1.963,
548
+ "step": 20745
549
+ },
550
+ {
551
+ "epoch": 9.11062906724512,
552
+ "grad_norm": 20100.53515625,
553
+ "learning_rate": 9.442591177115253e-05,
554
+ "loss": 0.973,
555
+ "step": 21000
556
+ },
557
+ {
558
+ "epoch": 9.327548806941431,
559
+ "grad_norm": 4432.14404296875,
560
+ "learning_rate": 9.416247841751377e-05,
561
+ "loss": 0.9844,
562
+ "step": 21500
563
+ },
564
+ {
565
+ "epoch": 9.544468546637743,
566
+ "grad_norm": 5448.53369140625,
567
+ "learning_rate": 9.389334809084711e-05,
568
+ "loss": 0.9783,
569
+ "step": 22000
570
+ },
571
+ {
572
+ "epoch": 9.761388286334057,
573
+ "grad_norm": 18744.869140625,
574
+ "learning_rate": 9.361855550904582e-05,
575
+ "loss": 0.9591,
576
+ "step": 22500
577
+ },
578
+ {
579
+ "epoch": 9.97830802603037,
580
+ "grad_norm": 9256.87890625,
581
+ "learning_rate": 9.333870255084278e-05,
582
+ "loss": 0.9536,
583
+ "step": 23000
584
+ },
585
+ {
586
+ "epoch": 10.0,
587
+ "eval_loss": 0.9346117377281189,
588
+ "eval_map": 0.4603,
589
+ "eval_map_50": 0.8028,
590
+ "eval_map_75": 0.4763,
591
+ "eval_map_bus": 0.5317,
592
+ "eval_map_car": 0.4808,
593
+ "eval_map_container": 0.5064,
594
+ "eval_map_large": 0.6826,
595
+ "eval_map_medium": 0.4525,
596
+ "eval_map_motorbike": 0.3221,
597
+ "eval_map_small": 0.108,
598
+ "eval_mar_1": 0.2479,
599
+ "eval_mar_10": 0.532,
600
+ "eval_mar_100": 0.5823,
601
+ "eval_mar_100_bus": 0.6637,
602
+ "eval_mar_100_car": 0.591,
603
+ "eval_mar_100_container": 0.6308,
604
+ "eval_mar_100_motorbike": 0.4437,
605
+ "eval_mar_large": 0.7735,
606
+ "eval_mar_medium": 0.5844,
607
+ "eval_mar_small": 0.2885,
608
+ "eval_runtime": 144.7391,
609
+ "eval_samples_per_second": 15.911,
610
+ "eval_steps_per_second": 1.99,
611
+ "step": 23050
612
+ },
613
+ {
614
+ "epoch": 10.195227765726681,
615
+ "grad_norm": 6654.49365234375,
616
+ "learning_rate": 9.30532812271677e-05,
617
+ "loss": 0.9821,
618
+ "step": 23500
619
+ },
620
+ {
621
+ "epoch": 10.412147505422993,
622
+ "grad_norm": 12540.853515625,
623
+ "learning_rate": 9.276173960919273e-05,
624
+ "loss": 0.9966,
625
+ "step": 24000
626
+ },
627
+ {
628
+ "epoch": 10.629067245119305,
629
+ "grad_norm": 8583.029296875,
630
+ "learning_rate": 9.246468171391835e-05,
631
+ "loss": 0.9826,
632
+ "step": 24500
633
+ },
634
+ {
635
+ "epoch": 10.845986984815617,
636
+ "grad_norm": 5945.16796875,
637
+ "learning_rate": 9.216214586190261e-05,
638
+ "loss": 0.9769,
639
+ "step": 25000
640
+ },
641
+ {
642
+ "epoch": 11.0,
643
+ "eval_loss": 0.9336058497428894,
644
+ "eval_map": 0.4511,
645
+ "eval_map_50": 0.8,
646
+ "eval_map_75": 0.4575,
647
+ "eval_map_bus": 0.5073,
648
+ "eval_map_car": 0.4708,
649
+ "eval_map_container": 0.5006,
650
+ "eval_map_large": 0.6611,
651
+ "eval_map_medium": 0.4469,
652
+ "eval_map_motorbike": 0.3259,
653
+ "eval_map_small": 0.1072,
654
+ "eval_mar_1": 0.2412,
655
+ "eval_mar_10": 0.5266,
656
+ "eval_mar_100": 0.5752,
657
+ "eval_mar_100_bus": 0.6493,
658
+ "eval_mar_100_car": 0.5772,
659
+ "eval_mar_100_container": 0.6233,
660
+ "eval_mar_100_motorbike": 0.4511,
661
+ "eval_mar_large": 0.7508,
662
+ "eval_mar_medium": 0.5789,
663
+ "eval_mar_small": 0.2908,
664
+ "eval_runtime": 147.6126,
665
+ "eval_samples_per_second": 15.602,
666
+ "eval_steps_per_second": 1.951,
667
+ "step": 25355
668
+ },
669
+ {
670
+ "epoch": 11.062906724511931,
671
+ "grad_norm": 10052.8896484375,
672
+ "learning_rate": 9.185417108036163e-05,
673
+ "loss": 0.9731,
674
+ "step": 25500
675
+ },
676
+ {
677
+ "epoch": 11.279826464208243,
678
+ "grad_norm": 3725.647216796875,
679
+ "learning_rate": 9.154079709813505e-05,
680
+ "loss": 0.9572,
681
+ "step": 26000
682
+ },
683
+ {
684
+ "epoch": 11.496746203904555,
685
+ "grad_norm": 10829.8896484375,
686
+ "learning_rate": 9.122206434056104e-05,
687
+ "loss": 0.9558,
688
+ "step": 26500
689
+ },
690
+ {
691
+ "epoch": 11.713665943600867,
692
+ "grad_norm": 16212.732421875,
693
+ "learning_rate": 9.08980139242614e-05,
694
+ "loss": 0.9622,
695
+ "step": 27000
696
+ },
697
+ {
698
+ "epoch": 11.93058568329718,
699
+ "grad_norm": 12326.623046875,
700
+ "learning_rate": 9.056868765183752e-05,
701
+ "loss": 0.9506,
702
+ "step": 27500
703
+ },
704
+ {
705
+ "epoch": 12.0,
706
+ "eval_loss": 0.9213413596153259,
707
+ "eval_map": 0.4617,
708
+ "eval_map_50": 0.8095,
709
+ "eval_map_75": 0.4735,
710
+ "eval_map_bus": 0.5252,
711
+ "eval_map_car": 0.4891,
712
+ "eval_map_container": 0.5052,
713
+ "eval_map_large": 0.676,
714
+ "eval_map_medium": 0.455,
715
+ "eval_map_motorbike": 0.3272,
716
+ "eval_map_small": 0.1083,
717
+ "eval_mar_1": 0.2461,
718
+ "eval_mar_10": 0.5327,
719
+ "eval_mar_100": 0.5818,
720
+ "eval_mar_100_bus": 0.6626,
721
+ "eval_mar_100_car": 0.5902,
722
+ "eval_mar_100_container": 0.627,
723
+ "eval_mar_100_motorbike": 0.4475,
724
+ "eval_mar_large": 0.7631,
725
+ "eval_mar_medium": 0.5838,
726
+ "eval_mar_small": 0.2935,
727
+ "eval_runtime": 143.9851,
728
+ "eval_samples_per_second": 15.995,
729
+ "eval_steps_per_second": 2.0,
730
+ "step": 27660
731
+ },
732
+ {
733
+ "epoch": 12.147505422993492,
734
+ "grad_norm": 23455.04296875,
735
+ "learning_rate": 9.023412800647782e-05,
736
+ "loss": 0.9338,
737
+ "step": 28000
738
+ },
739
+ {
740
+ "epoch": 12.364425162689805,
741
+ "grad_norm": 7761.9912109375,
742
+ "learning_rate": 8.989506279701155e-05,
743
+ "loss": 0.9535,
744
+ "step": 28500
745
+ },
746
+ {
747
+ "epoch": 12.581344902386117,
748
+ "grad_norm": 7034.04052734375,
749
+ "learning_rate": 8.955017679879593e-05,
750
+ "loss": 0.9507,
751
+ "step": 29000
752
+ },
753
+ {
754
+ "epoch": 12.79826464208243,
755
+ "grad_norm": 5548.38525390625,
756
+ "learning_rate": 8.920018881585156e-05,
757
+ "loss": 0.9417,
758
+ "step": 29500
759
+ },
760
+ {
761
+ "epoch": 13.0,
762
+ "eval_loss": 0.9276334643363953,
763
+ "eval_map": 0.4608,
764
+ "eval_map_50": 0.8076,
765
+ "eval_map_75": 0.4687,
766
+ "eval_map_bus": 0.5336,
767
+ "eval_map_car": 0.4866,
768
+ "eval_map_container": 0.4992,
769
+ "eval_map_large": 0.6724,
770
+ "eval_map_medium": 0.4511,
771
+ "eval_map_motorbike": 0.3238,
772
+ "eval_map_small": 0.1064,
773
+ "eval_mar_1": 0.2472,
774
+ "eval_mar_10": 0.5288,
775
+ "eval_mar_100": 0.5752,
776
+ "eval_mar_100_bus": 0.6588,
777
+ "eval_mar_100_car": 0.5895,
778
+ "eval_mar_100_container": 0.6158,
779
+ "eval_mar_100_motorbike": 0.4369,
780
+ "eval_mar_large": 0.7512,
781
+ "eval_mar_medium": 0.5708,
782
+ "eval_mar_small": 0.2981,
783
+ "eval_runtime": 145.1236,
784
+ "eval_samples_per_second": 15.869,
785
+ "eval_steps_per_second": 1.985,
786
+ "step": 29965
787
+ },
788
+ {
789
+ "epoch": 13.015184381778742,
790
+ "grad_norm": 12253.849609375,
791
+ "learning_rate": 8.884514399673369e-05,
792
+ "loss": 0.9555,
793
+ "step": 30000
794
+ },
795
+ {
796
+ "epoch": 13.232104121475054,
797
+ "grad_norm": 13796.53125,
798
+ "learning_rate": 8.848508814233106e-05,
799
+ "loss": 0.9476,
800
+ "step": 30500
801
+ },
802
+ {
803
+ "epoch": 13.449023861171366,
804
+ "grad_norm": 5713.0361328125,
805
+ "learning_rate": 8.812080266430921e-05,
806
+ "loss": 0.9417,
807
+ "step": 31000
808
+ },
809
+ {
810
+ "epoch": 13.66594360086768,
811
+ "grad_norm": 3508.00634765625,
812
+ "learning_rate": 8.77508745092958e-05,
813
+ "loss": 0.9446,
814
+ "step": 31500
815
+ },
816
+ {
817
+ "epoch": 13.882863340563992,
818
+ "grad_norm": 4413.595703125,
819
+ "learning_rate": 8.737607648009131e-05,
820
+ "loss": 0.9488,
821
+ "step": 32000
822
+ },
823
+ {
824
+ "epoch": 14.0,
825
+ "eval_loss": 0.9241105318069458,
826
+ "eval_map": 0.4602,
827
+ "eval_map_50": 0.8064,
828
+ "eval_map_75": 0.4696,
829
+ "eval_map_bus": 0.5309,
830
+ "eval_map_car": 0.4761,
831
+ "eval_map_container": 0.5062,
832
+ "eval_map_large": 0.6817,
833
+ "eval_map_medium": 0.4503,
834
+ "eval_map_motorbike": 0.3275,
835
+ "eval_map_small": 0.1122,
836
+ "eval_mar_1": 0.2462,
837
+ "eval_mar_10": 0.5333,
838
+ "eval_mar_100": 0.5819,
839
+ "eval_mar_100_bus": 0.6633,
840
+ "eval_mar_100_car": 0.5828,
841
+ "eval_mar_100_container": 0.6259,
842
+ "eval_mar_100_motorbike": 0.4557,
843
+ "eval_mar_large": 0.7655,
844
+ "eval_mar_medium": 0.5804,
845
+ "eval_mar_small": 0.2896,
846
+ "eval_runtime": 145.7229,
847
+ "eval_samples_per_second": 15.804,
848
+ "eval_steps_per_second": 1.976,
849
+ "step": 32270
850
+ },
851
+ {
852
+ "epoch": 14.099783080260304,
853
+ "grad_norm": 7677.90478515625,
854
+ "learning_rate": 8.699645692575454e-05,
855
+ "loss": 0.9519,
856
+ "step": 32500
857
+ },
858
+ {
859
+ "epoch": 14.316702819956616,
860
+ "grad_norm": 4366.18212890625,
861
+ "learning_rate": 8.661283833169138e-05,
862
+ "loss": 0.9471,
863
+ "step": 33000
864
+ },
865
+ {
866
+ "epoch": 14.533622559652928,
867
+ "grad_norm": 5411.373046875,
868
+ "learning_rate": 8.622373265190294e-05,
869
+ "loss": 0.9423,
870
+ "step": 33500
871
+ },
872
+ {
873
+ "epoch": 14.750542299349242,
874
+ "grad_norm": 7791.236328125,
875
+ "learning_rate": 8.582995409968335e-05,
876
+ "loss": 0.9357,
877
+ "step": 34000
878
+ },
879
+ {
880
+ "epoch": 14.967462039045554,
881
+ "grad_norm": 20878.41796875,
882
+ "learning_rate": 8.543155347258457e-05,
883
+ "loss": 0.9371,
884
+ "step": 34500
885
+ },
886
+ {
887
+ "epoch": 15.0,
888
+ "eval_loss": 0.9158653020858765,
889
+ "eval_map": 0.4646,
890
+ "eval_map_50": 0.8109,
891
+ "eval_map_75": 0.479,
892
+ "eval_map_bus": 0.525,
893
+ "eval_map_car": 0.4942,
894
+ "eval_map_container": 0.5118,
895
+ "eval_map_large": 0.6799,
896
+ "eval_map_medium": 0.4567,
897
+ "eval_map_motorbike": 0.3276,
898
+ "eval_map_small": 0.1141,
899
+ "eval_mar_1": 0.2484,
900
+ "eval_mar_10": 0.533,
901
+ "eval_mar_100": 0.5795,
902
+ "eval_mar_100_bus": 0.6523,
903
+ "eval_mar_100_car": 0.5909,
904
+ "eval_mar_100_container": 0.6278,
905
+ "eval_mar_100_motorbike": 0.447,
906
+ "eval_mar_large": 0.7745,
907
+ "eval_mar_medium": 0.5767,
908
+ "eval_mar_small": 0.3179,
909
+ "eval_runtime": 143.8949,
910
+ "eval_samples_per_second": 16.005,
911
+ "eval_steps_per_second": 2.001,
912
+ "step": 34575
913
+ },
914
+ {
915
+ "epoch": 15.184381778741866,
916
+ "grad_norm": 8804.869140625,
917
+ "learning_rate": 8.502858216440758e-05,
918
+ "loss": 0.9249,
919
+ "step": 35000
920
+ },
921
+ {
922
+ "epoch": 15.401301518438178,
923
+ "grad_norm": 7941.79443359375,
924
+ "learning_rate": 8.462191161340177e-05,
925
+ "loss": 0.9341,
926
+ "step": 35500
927
+ },
928
+ {
929
+ "epoch": 15.61822125813449,
930
+ "grad_norm": 11279.1513671875,
931
+ "learning_rate": 8.420996435565979e-05,
932
+ "loss": 0.9342,
933
+ "step": 36000
934
+ },
935
+ {
936
+ "epoch": 15.835140997830802,
937
+ "grad_norm": 2519.469970703125,
938
+ "learning_rate": 8.379360400220391e-05,
939
+ "loss": 0.9311,
940
+ "step": 36500
941
+ },
942
+ {
943
+ "epoch": 16.0,
944
+ "eval_loss": 0.8999801278114319,
945
+ "eval_map": 0.4741,
946
+ "eval_map_50": 0.8169,
947
+ "eval_map_75": 0.4893,
948
+ "eval_map_bus": 0.5392,
949
+ "eval_map_car": 0.5006,
950
+ "eval_map_container": 0.5198,
951
+ "eval_map_large": 0.6914,
952
+ "eval_map_medium": 0.4673,
953
+ "eval_map_motorbike": 0.3368,
954
+ "eval_map_small": 0.1224,
955
+ "eval_mar_1": 0.2523,
956
+ "eval_mar_10": 0.544,
957
+ "eval_mar_100": 0.5909,
958
+ "eval_mar_100_bus": 0.6792,
959
+ "eval_mar_100_car": 0.5917,
960
+ "eval_mar_100_container": 0.6356,
961
+ "eval_mar_100_motorbike": 0.4571,
962
+ "eval_mar_large": 0.7817,
963
+ "eval_mar_medium": 0.593,
964
+ "eval_mar_small": 0.3192,
965
+ "eval_runtime": 144.4917,
966
+ "eval_samples_per_second": 15.939,
967
+ "eval_steps_per_second": 1.993,
968
+ "step": 36880
969
+ },
970
+ {
971
+ "epoch": 16.052060737527114,
972
+ "grad_norm": 6870.65673828125,
973
+ "learning_rate": 8.337288426364522e-05,
974
+ "loss": 0.9394,
975
+ "step": 37000
976
+ },
977
+ {
978
+ "epoch": 16.268980477223426,
979
+ "grad_norm": 18748.70703125,
980
+ "learning_rate": 8.294785941295685e-05,
981
+ "loss": 0.9206,
982
+ "step": 37500
983
+ },
984
+ {
985
+ "epoch": 16.485900216919738,
986
+ "grad_norm": 8770.451171875,
987
+ "learning_rate": 8.251944703380834e-05,
988
+ "loss": 0.9277,
989
+ "step": 38000
990
+ },
991
+ {
992
+ "epoch": 16.702819956616054,
993
+ "grad_norm": 10527.71484375,
994
+ "learning_rate": 8.208598532633761e-05,
995
+ "loss": 0.9363,
996
+ "step": 38500
997
+ },
998
+ {
999
+ "epoch": 16.919739696312366,
1000
+ "grad_norm": 13457.677734375,
1001
+ "learning_rate": 8.164838451708801e-05,
1002
+ "loss": 0.9284,
1003
+ "step": 39000
1004
+ },
1005
+ {
1006
+ "epoch": 17.0,
1007
+ "eval_loss": 0.9076481461524963,
1008
+ "eval_map": 0.472,
1009
+ "eval_map_50": 0.8148,
1010
+ "eval_map_75": 0.4845,
1011
+ "eval_map_bus": 0.5397,
1012
+ "eval_map_car": 0.4969,
1013
+ "eval_map_container": 0.5162,
1014
+ "eval_map_large": 0.684,
1015
+ "eval_map_medium": 0.4625,
1016
+ "eval_map_motorbike": 0.3354,
1017
+ "eval_map_small": 0.1174,
1018
+ "eval_mar_1": 0.252,
1019
+ "eval_mar_10": 0.5415,
1020
+ "eval_mar_100": 0.5905,
1021
+ "eval_mar_100_bus": 0.6732,
1022
+ "eval_mar_100_car": 0.5938,
1023
+ "eval_mar_100_container": 0.6354,
1024
+ "eval_mar_100_motorbike": 0.4596,
1025
+ "eval_mar_large": 0.767,
1026
+ "eval_mar_medium": 0.5905,
1027
+ "eval_mar_small": 0.315,
1028
+ "eval_runtime": 145.0152,
1029
+ "eval_samples_per_second": 15.881,
1030
+ "eval_steps_per_second": 1.986,
1031
+ "step": 39185
1032
+ },
1033
+ {
1034
+ "epoch": 17.136659436008678,
1035
+ "grad_norm": 6826.60595703125,
1036
+ "learning_rate": 8.120670105669583e-05,
1037
+ "loss": 0.9237,
1038
+ "step": 39500
1039
+ },
1040
+ {
1041
+ "epoch": 17.35357917570499,
1042
+ "grad_norm": 10753.5400390625,
1043
+ "learning_rate": 8.076099192246061e-05,
1044
+ "loss": 0.9257,
1045
+ "step": 40000
1046
+ },
1047
+ {
1048
+ "epoch": 17.570498915401302,
1049
+ "grad_norm": 5758.92919921875,
1050
+ "learning_rate": 8.031131461099501e-05,
1051
+ "loss": 0.9247,
1052
+ "step": 40500
1053
+ },
1054
+ {
1055
+ "epoch": 17.787418655097614,
1056
+ "grad_norm": 11779.173828125,
1057
+ "learning_rate": 7.985772713080777e-05,
1058
+ "loss": 0.9107,
1059
+ "step": 41000
1060
+ },
1061
+ {
1062
+ "epoch": 18.0,
1063
+ "eval_loss": 0.9001684188842773,
1064
+ "eval_map": 0.4753,
1065
+ "eval_map_50": 0.8198,
1066
+ "eval_map_75": 0.4952,
1067
+ "eval_map_bus": 0.5451,
1068
+ "eval_map_car": 0.501,
1069
+ "eval_map_container": 0.5198,
1070
+ "eval_map_large": 0.6944,
1071
+ "eval_map_medium": 0.4679,
1072
+ "eval_map_motorbike": 0.3352,
1073
+ "eval_map_small": 0.1207,
1074
+ "eval_mar_1": 0.2516,
1075
+ "eval_mar_10": 0.5428,
1076
+ "eval_mar_100": 0.5902,
1077
+ "eval_mar_100_bus": 0.6716,
1078
+ "eval_mar_100_car": 0.5977,
1079
+ "eval_mar_100_container": 0.6353,
1080
+ "eval_mar_100_motorbike": 0.4562,
1081
+ "eval_mar_large": 0.7743,
1082
+ "eval_mar_medium": 0.592,
1083
+ "eval_mar_small": 0.3133,
1084
+ "eval_runtime": 143.7248,
1085
+ "eval_samples_per_second": 16.024,
1086
+ "eval_steps_per_second": 2.004,
1087
+ "step": 41490
1088
+ },
1089
+ {
1090
+ "epoch": 18.004338394793926,
1091
+ "grad_norm": 3542.92626953125,
1092
+ "learning_rate": 7.940120667790506e-05,
1093
+ "loss": 0.919,
1094
+ "step": 41500
1095
+ },
1096
+ {
1097
+ "epoch": 18.22125813449024,
1098
+ "grad_norm": 22165.880859375,
1099
+ "learning_rate": 7.893998242197806e-05,
1100
+ "loss": 0.928,
1101
+ "step": 42000
1102
+ },
1103
+ {
1104
+ "epoch": 18.43817787418655,
1105
+ "grad_norm": 3429.292236328125,
1106
+ "learning_rate": 7.84750248995952e-05,
1107
+ "loss": 0.9192,
1108
+ "step": 42500
1109
+ },
1110
+ {
1111
+ "epoch": 18.655097613882862,
1112
+ "grad_norm": 10154.87109375,
1113
+ "learning_rate": 7.800639409041694e-05,
1114
+ "loss": 0.9183,
1115
+ "step": 43000
1116
+ },
1117
+ {
1118
+ "epoch": 18.872017353579174,
1119
+ "grad_norm": 25265.55078125,
1120
+ "learning_rate": 7.753415044795898e-05,
1121
+ "loss": 0.9125,
1122
+ "step": 43500
1123
+ },
1124
+ {
1125
+ "epoch": 19.0,
1126
+ "eval_loss": 0.8902826905250549,
1127
+ "eval_map": 0.4805,
1128
+ "eval_map_50": 0.8223,
1129
+ "eval_map_75": 0.5006,
1130
+ "eval_map_bus": 0.5537,
1131
+ "eval_map_car": 0.5047,
1132
+ "eval_map_container": 0.5198,
1133
+ "eval_map_large": 0.6876,
1134
+ "eval_map_medium": 0.4736,
1135
+ "eval_map_motorbike": 0.3438,
1136
+ "eval_map_small": 0.119,
1137
+ "eval_mar_1": 0.2528,
1138
+ "eval_mar_10": 0.5449,
1139
+ "eval_mar_100": 0.5921,
1140
+ "eval_mar_100_bus": 0.6688,
1141
+ "eval_mar_100_car": 0.6008,
1142
+ "eval_mar_100_container": 0.6319,
1143
+ "eval_mar_100_motorbike": 0.4669,
1144
+ "eval_mar_large": 0.7716,
1145
+ "eval_mar_medium": 0.5905,
1146
+ "eval_mar_small": 0.3164,
1147
+ "eval_runtime": 146.715,
1148
+ "eval_samples_per_second": 15.697,
1149
+ "eval_steps_per_second": 1.963,
1150
+ "step": 43795
1151
+ },
1152
+ {
1153
+ "epoch": 19.088937093275486,
1154
+ "grad_norm": 7347.31640625,
1155
+ "learning_rate": 7.705835489179352e-05,
1156
+ "loss": 0.9033,
1157
+ "step": 44000
1158
+ },
1159
+ {
1160
+ "epoch": 19.305856832971802,
1161
+ "grad_norm": 5815.42822265625,
1162
+ "learning_rate": 7.658003081440886e-05,
1163
+ "loss": 0.9049,
1164
+ "step": 44500
1165
+ },
1166
+ {
1167
+ "epoch": 19.522776572668114,
1168
+ "grad_norm": 5560.4921875,
1169
+ "learning_rate": 7.609732280983515e-05,
1170
+ "loss": 0.9124,
1171
+ "step": 45000
1172
+ },
1173
+ {
1174
+ "epoch": 19.739696312364426,
1175
+ "grad_norm": 5130.99462890625,
1176
+ "learning_rate": 7.56112482427524e-05,
1177
+ "loss": 0.9205,
1178
+ "step": 45500
1179
+ },
1180
+ {
1181
+ "epoch": 19.95661605206074,
1182
+ "grad_norm": 17505.642578125,
1183
+ "learning_rate": 7.512186981692624e-05,
1184
+ "loss": 0.913,
1185
+ "step": 46000
1186
+ },
1187
+ {
1188
+ "epoch": 20.0,
1189
+ "eval_loss": 0.8936046957969666,
1190
+ "eval_map": 0.4822,
1191
+ "eval_map_50": 0.8243,
1192
+ "eval_map_75": 0.504,
1193
+ "eval_map_bus": 0.5592,
1194
+ "eval_map_car": 0.4984,
1195
+ "eval_map_container": 0.5236,
1196
+ "eval_map_large": 0.6904,
1197
+ "eval_map_medium": 0.4777,
1198
+ "eval_map_motorbike": 0.3475,
1199
+ "eval_map_small": 0.115,
1200
+ "eval_mar_1": 0.2536,
1201
+ "eval_mar_10": 0.5453,
1202
+ "eval_mar_100": 0.5923,
1203
+ "eval_mar_100_bus": 0.6727,
1204
+ "eval_mar_100_car": 0.5953,
1205
+ "eval_mar_100_container": 0.6347,
1206
+ "eval_mar_100_motorbike": 0.4666,
1207
+ "eval_mar_large": 0.7683,
1208
+ "eval_mar_medium": 0.5946,
1209
+ "eval_mar_small": 0.3124,
1210
+ "eval_runtime": 146.1172,
1211
+ "eval_samples_per_second": 15.761,
1212
+ "eval_steps_per_second": 1.971,
1213
+ "step": 46100
1214
+ },
1215
+ {
1216
+ "epoch": 20.17353579175705,
1217
+ "grad_norm": 21176.666015625,
1218
+ "learning_rate": 7.462925066232113e-05,
1219
+ "loss": 0.9039,
1220
+ "step": 46500
1221
+ },
1222
+ {
1223
+ "epoch": 20.390455531453362,
1224
+ "grad_norm": 4085.164306640625,
1225
+ "learning_rate": 7.413444904801033e-05,
1226
+ "loss": 0.9082,
1227
+ "step": 47000
1228
+ },
1229
+ {
1230
+ "epoch": 20.607375271149674,
1231
+ "grad_norm": 20771.896484375,
1232
+ "learning_rate": 7.363554565211042e-05,
1233
+ "loss": 0.9087,
1234
+ "step": 47500
1235
+ },
1236
+ {
1237
+ "epoch": 20.824295010845987,
1238
+ "grad_norm": 5990.69580078125,
1239
+ "learning_rate": 7.313359326369684e-05,
1240
+ "loss": 0.908,
1241
+ "step": 48000
1242
+ },
1243
+ {
1244
+ "epoch": 21.0,
1245
+ "eval_loss": 0.8810420036315918,
1246
+ "eval_map": 0.483,
1247
+ "eval_map_50": 0.8249,
1248
+ "eval_map_75": 0.5047,
1249
+ "eval_map_bus": 0.5531,
1250
+ "eval_map_car": 0.5081,
1251
+ "eval_map_container": 0.5229,
1252
+ "eval_map_large": 0.7091,
1253
+ "eval_map_medium": 0.4743,
1254
+ "eval_map_motorbike": 0.348,
1255
+ "eval_map_small": 0.1233,
1256
+ "eval_mar_1": 0.2539,
1257
+ "eval_mar_10": 0.5486,
1258
+ "eval_mar_100": 0.5987,
1259
+ "eval_mar_100_bus": 0.6764,
1260
+ "eval_mar_100_car": 0.6079,
1261
+ "eval_mar_100_container": 0.6426,
1262
+ "eval_mar_100_motorbike": 0.4679,
1263
+ "eval_mar_large": 0.7947,
1264
+ "eval_mar_medium": 0.5992,
1265
+ "eval_mar_small": 0.3117,
1266
+ "eval_runtime": 143.798,
1267
+ "eval_samples_per_second": 16.016,
1268
+ "eval_steps_per_second": 2.003,
1269
+ "step": 48405
1270
+ },
1271
+ {
1272
+ "epoch": 21.0412147505423,
1273
+ "grad_norm": 4679.26220703125,
1274
+ "learning_rate": 7.262865663477902e-05,
1275
+ "loss": 0.9023,
1276
+ "step": 48500
1277
+ },
1278
+ {
1279
+ "epoch": 21.25813449023861,
1280
+ "grad_norm": 18654.177734375,
1281
+ "learning_rate": 7.21208009023343e-05,
1282
+ "loss": 0.9069,
1283
+ "step": 49000
1284
+ },
1285
+ {
1286
+ "epoch": 21.475054229934923,
1287
+ "grad_norm": 10939.3623046875,
1288
+ "learning_rate": 7.161009157990521e-05,
1289
+ "loss": 0.8956,
1290
+ "step": 49500
1291
+ },
1292
+ {
1293
+ "epoch": 21.691973969631235,
1294
+ "grad_norm": 14381.9765625,
1295
+ "learning_rate": 7.109659454914833e-05,
1296
+ "loss": 0.8925,
1297
+ "step": 50000
1298
+ },
1299
+ {
1300
+ "epoch": 21.90889370932755,
1301
+ "grad_norm": 5005.86865234375,
1302
+ "learning_rate": 7.058141116015095e-05,
1303
+ "loss": 0.8997,
1304
+ "step": 50500
1305
+ },
1306
+ {
1307
+ "epoch": 22.0,
1308
+ "eval_loss": 0.8783358335494995,
1309
+ "eval_map": 0.4846,
1310
+ "eval_map_50": 0.8275,
1311
+ "eval_map_75": 0.5024,
1312
+ "eval_map_bus": 0.5507,
1313
+ "eval_map_car": 0.5086,
1314
+ "eval_map_container": 0.5299,
1315
+ "eval_map_large": 0.699,
1316
+ "eval_map_medium": 0.4761,
1317
+ "eval_map_motorbike": 0.3491,
1318
+ "eval_map_small": 0.1256,
1319
+ "eval_mar_1": 0.2542,
1320
+ "eval_mar_10": 0.5503,
1321
+ "eval_mar_100": 0.5994,
1322
+ "eval_mar_100_bus": 0.6731,
1323
+ "eval_mar_100_car": 0.6068,
1324
+ "eval_mar_100_container": 0.6486,
1325
+ "eval_mar_100_motorbike": 0.469,
1326
+ "eval_mar_large": 0.7854,
1327
+ "eval_mar_medium": 0.5993,
1328
+ "eval_mar_small": 0.3314,
1329
+ "eval_runtime": 144.6256,
1330
+ "eval_samples_per_second": 15.924,
1331
+ "eval_steps_per_second": 1.991,
1332
+ "step": 50710
1333
+ },
1334
+ {
1335
+ "epoch": 22.125813449023862,
1336
+ "grad_norm": 19755.140625,
1337
+ "learning_rate": 7.006254303068497e-05,
1338
+ "loss": 0.8975,
1339
+ "step": 51000
1340
+ },
1341
+ {
1342
+ "epoch": 22.342733188720175,
1343
+ "grad_norm": 11582.6494140625,
1344
+ "learning_rate": 6.954108682712056e-05,
1345
+ "loss": 0.898,
1346
+ "step": 51500
1347
+ },
1348
+ {
1349
+ "epoch": 22.559652928416487,
1350
+ "grad_norm": 6819.56396484375,
1351
+ "learning_rate": 6.901710981746509e-05,
1352
+ "loss": 0.8999,
1353
+ "step": 52000
1354
+ },
1355
+ {
1356
+ "epoch": 22.7765726681128,
1357
+ "grad_norm": 7609.8115234375,
1358
+ "learning_rate": 6.849067959491079e-05,
1359
+ "loss": 0.8978,
1360
+ "step": 52500
1361
+ },
1362
+ {
1363
+ "epoch": 22.99349240780911,
1364
+ "grad_norm": 10856.244140625,
1365
+ "learning_rate": 6.796186406911505e-05,
1366
+ "loss": 0.8865,
1367
+ "step": 53000
1368
+ },
1369
+ {
1370
+ "epoch": 23.0,
1371
+ "eval_loss": 0.871816098690033,
1372
+ "eval_map": 0.4901,
1373
+ "eval_map_50": 0.8289,
1374
+ "eval_map_75": 0.5202,
1375
+ "eval_map_bus": 0.5591,
1376
+ "eval_map_car": 0.5153,
1377
+ "eval_map_container": 0.5306,
1378
+ "eval_map_large": 0.7037,
1379
+ "eval_map_medium": 0.4841,
1380
+ "eval_map_motorbike": 0.3553,
1381
+ "eval_map_small": 0.1243,
1382
+ "eval_mar_1": 0.2558,
1383
+ "eval_mar_10": 0.5526,
1384
+ "eval_mar_100": 0.6008,
1385
+ "eval_mar_100_bus": 0.6757,
1386
+ "eval_mar_100_car": 0.6077,
1387
+ "eval_mar_100_container": 0.6452,
1388
+ "eval_mar_100_motorbike": 0.4746,
1389
+ "eval_mar_large": 0.788,
1390
+ "eval_mar_medium": 0.6019,
1391
+ "eval_mar_small": 0.3201,
1392
+ "eval_runtime": 145.8422,
1393
+ "eval_samples_per_second": 15.791,
1394
+ "eval_steps_per_second": 1.975,
1395
+ "step": 53015
1396
+ },
1397
+ {
1398
+ "epoch": 23.210412147505423,
1399
+ "grad_norm": 6373.787109375,
1400
+ "learning_rate": 6.743286051281191e-05,
1401
+ "loss": 0.8974,
1402
+ "step": 53500
1403
+ },
1404
+ {
1405
+ "epoch": 23.427331887201735,
1406
+ "grad_norm": 3711.38671875,
1407
+ "learning_rate": 6.68994881889314e-05,
1408
+ "loss": 0.9048,
1409
+ "step": 54000
1410
+ },
1411
+ {
1412
+ "epoch": 23.644251626898047,
1413
+ "grad_norm": 1683.364501953125,
1414
+ "learning_rate": 6.636393582598046e-05,
1415
+ "loss": 0.911,
1416
+ "step": 54500
1417
+ },
1418
+ {
1419
+ "epoch": 23.86117136659436,
1420
+ "grad_norm": 2381.84423828125,
1421
+ "learning_rate": 6.582627251037533e-05,
1422
+ "loss": 0.9052,
1423
+ "step": 55000
1424
+ },
1425
+ {
1426
+ "epoch": 24.0,
1427
+ "eval_loss": 0.8951303958892822,
1428
+ "eval_map": 0.4779,
1429
+ "eval_map_50": 0.8254,
1430
+ "eval_map_75": 0.4914,
1431
+ "eval_map_bus": 0.5509,
1432
+ "eval_map_car": 0.5002,
1433
+ "eval_map_container": 0.5227,
1434
+ "eval_map_large": 0.697,
1435
+ "eval_map_medium": 0.469,
1436
+ "eval_map_motorbike": 0.3376,
1437
+ "eval_map_small": 0.1198,
1438
+ "eval_mar_1": 0.2524,
1439
+ "eval_mar_10": 0.5426,
1440
+ "eval_mar_100": 0.5902,
1441
+ "eval_mar_100_bus": 0.6715,
1442
+ "eval_mar_100_car": 0.5914,
1443
+ "eval_mar_100_container": 0.643,
1444
+ "eval_mar_100_motorbike": 0.455,
1445
+ "eval_mar_large": 0.7785,
1446
+ "eval_mar_medium": 0.5885,
1447
+ "eval_mar_small": 0.319,
1448
+ "eval_runtime": 145.8551,
1449
+ "eval_samples_per_second": 15.79,
1450
+ "eval_steps_per_second": 1.975,
1451
+ "step": 55320
1452
+ },
1453
+ {
1454
+ "epoch": 24.07809110629067,
1455
+ "grad_norm": 19113.990234375,
1456
+ "learning_rate": 6.528656760084578e-05,
1457
+ "loss": 0.9112,
1458
+ "step": 55500
1459
+ },
1460
+ {
1461
+ "epoch": 24.295010845986983,
1462
+ "grad_norm": 20675.185546875,
1463
+ "learning_rate": 6.474489071948787e-05,
1464
+ "loss": 0.8944,
1465
+ "step": 56000
1466
+ },
1467
+ {
1468
+ "epoch": 24.5119305856833,
1469
+ "grad_norm": 10454.0859375,
1470
+ "learning_rate": 6.420131174278254e-05,
1471
+ "loss": 0.8915,
1472
+ "step": 56500
1473
+ },
1474
+ {
1475
+ "epoch": 24.72885032537961,
1476
+ "grad_norm": 4770.7431640625,
1477
+ "learning_rate": 6.365590079258165e-05,
1478
+ "loss": 0.8891,
1479
+ "step": 57000
1480
+ },
1481
+ {
1482
+ "epoch": 24.945770065075923,
1483
+ "grad_norm": 16598.36328125,
1484
+ "learning_rate": 6.310872822706217e-05,
1485
+ "loss": 0.8982,
1486
+ "step": 57500
1487
+ },
1488
+ {
1489
+ "epoch": 25.0,
1490
+ "eval_loss": 0.8864476084709167,
1491
+ "eval_map": 0.4662,
1492
+ "eval_map_50": 0.8105,
1493
+ "eval_map_75": 0.4884,
1494
+ "eval_map_bus": 0.515,
1495
+ "eval_map_car": 0.4889,
1496
+ "eval_map_container": 0.5084,
1497
+ "eval_map_large": 0.6737,
1498
+ "eval_map_medium": 0.4711,
1499
+ "eval_map_motorbike": 0.3524,
1500
+ "eval_map_small": 0.1286,
1501
+ "eval_mar_1": 0.2482,
1502
+ "eval_mar_10": 0.5444,
1503
+ "eval_mar_100": 0.5934,
1504
+ "eval_mar_100_bus": 0.6696,
1505
+ "eval_mar_100_car": 0.5993,
1506
+ "eval_mar_100_container": 0.6371,
1507
+ "eval_mar_100_motorbike": 0.4675,
1508
+ "eval_mar_large": 0.7729,
1509
+ "eval_mar_medium": 0.5969,
1510
+ "eval_mar_small": 0.3226,
1511
+ "eval_runtime": 144.7854,
1512
+ "eval_samples_per_second": 15.906,
1513
+ "eval_steps_per_second": 1.989,
1514
+ "step": 57625
1515
+ },
1516
+ {
1517
+ "epoch": 25.162689804772235,
1518
+ "grad_norm": 6422.17041015625,
1519
+ "learning_rate": 6.25609639994632e-05,
1520
+ "loss": 0.8784,
1521
+ "step": 58000
1522
+ },
1523
+ {
1524
+ "epoch": 25.379609544468547,
1525
+ "grad_norm": 16689.212890625,
1526
+ "learning_rate": 6.201048334737784e-05,
1527
+ "loss": 0.8855,
1528
+ "step": 58500
1529
+ },
1530
+ {
1531
+ "epoch": 25.59652928416486,
1532
+ "grad_norm": 18894.322265625,
1533
+ "learning_rate": 6.145845333932029e-05,
1534
+ "loss": 0.8872,
1535
+ "step": 59000
1536
+ },
1537
+ {
1538
+ "epoch": 25.81344902386117,
1539
+ "grad_norm": 4359.09521484375,
1540
+ "learning_rate": 6.090494518732801e-05,
1541
+ "loss": 0.8835,
1542
+ "step": 59500
1543
+ },
1544
+ {
1545
+ "epoch": 26.0,
1546
+ "eval_loss": 0.8596058487892151,
1547
+ "eval_map": 0.4972,
1548
+ "eval_map_50": 0.8376,
1549
+ "eval_map_75": 0.5257,
1550
+ "eval_map_bus": 0.5714,
1551
+ "eval_map_car": 0.5186,
1552
+ "eval_map_container": 0.5391,
1553
+ "eval_map_large": 0.7028,
1554
+ "eval_map_medium": 0.4915,
1555
+ "eval_map_motorbike": 0.3598,
1556
+ "eval_map_small": 0.1342,
1557
+ "eval_mar_1": 0.2583,
1558
+ "eval_mar_10": 0.5574,
1559
+ "eval_mar_100": 0.6057,
1560
+ "eval_mar_100_bus": 0.6883,
1561
+ "eval_mar_100_car": 0.6094,
1562
+ "eval_mar_100_container": 0.6514,
1563
+ "eval_mar_100_motorbike": 0.4737,
1564
+ "eval_mar_large": 0.7866,
1565
+ "eval_mar_medium": 0.6078,
1566
+ "eval_mar_small": 0.3419,
1567
+ "eval_runtime": 145.0724,
1568
+ "eval_samples_per_second": 15.875,
1569
+ "eval_steps_per_second": 1.985,
1570
+ "step": 59930
1571
+ },
1572
+ {
1573
+ "epoch": 26.030368763557483,
1574
+ "grad_norm": 28658.181640625,
1575
+ "learning_rate": 6.0350030294119464e-05,
1576
+ "loss": 0.8825,
1577
+ "step": 60000
1578
+ },
1579
+ {
1580
+ "epoch": 26.247288503253795,
1581
+ "grad_norm": 6283.97509765625,
1582
+ "learning_rate": 5.979489402880387e-05,
1583
+ "loss": 0.8833,
1584
+ "step": 60500
1585
+ },
1586
+ {
1587
+ "epoch": 26.464208242950107,
1588
+ "grad_norm": 4272.13037109375,
1589
+ "learning_rate": 5.923738303304234e-05,
1590
+ "loss": 0.8844,
1591
+ "step": 61000
1592
+ },
1593
+ {
1594
+ "epoch": 26.68112798264642,
1595
+ "grad_norm": 14047.6572265625,
1596
+ "learning_rate": 5.8678680412085076e-05,
1597
+ "loss": 0.8785,
1598
+ "step": 61500
1599
+ },
1600
+ {
1601
+ "epoch": 26.89804772234273,
1602
+ "grad_norm": 8514.8544921875,
1603
+ "learning_rate": 5.811885823873863e-05,
1604
+ "loss": 0.8692,
1605
+ "step": 62000
1606
+ },
1607
+ {
1608
+ "epoch": 27.0,
1609
+ "eval_loss": 0.8606675863265991,
1610
+ "eval_map": 0.4927,
1611
+ "eval_map_50": 0.837,
1612
+ "eval_map_75": 0.5186,
1613
+ "eval_map_bus": 0.5641,
1614
+ "eval_map_car": 0.5122,
1615
+ "eval_map_container": 0.5369,
1616
+ "eval_map_large": 0.7052,
1617
+ "eval_map_medium": 0.4874,
1618
+ "eval_map_motorbike": 0.3576,
1619
+ "eval_map_small": 0.1426,
1620
+ "eval_mar_1": 0.2578,
1621
+ "eval_mar_10": 0.5566,
1622
+ "eval_mar_100": 0.6043,
1623
+ "eval_mar_100_bus": 0.6885,
1624
+ "eval_mar_100_car": 0.6043,
1625
+ "eval_mar_100_container": 0.6507,
1626
+ "eval_mar_100_motorbike": 0.4737,
1627
+ "eval_mar_large": 0.7936,
1628
+ "eval_mar_medium": 0.608,
1629
+ "eval_mar_small": 0.3251,
1630
+ "eval_runtime": 145.9029,
1631
+ "eval_samples_per_second": 15.784,
1632
+ "eval_steps_per_second": 1.974,
1633
+ "step": 62235
1634
+ },
1635
+ {
1636
+ "epoch": 27.114967462039047,
1637
+ "grad_norm": 17592.478515625,
1638
+ "learning_rate": 5.755798873023209e-05,
1639
+ "loss": 0.8719,
1640
+ "step": 62500
1641
+ },
1642
+ {
1643
+ "epoch": 27.33188720173536,
1644
+ "grad_norm": 28705.041015625,
1645
+ "learning_rate": 5.699614423890112e-05,
1646
+ "loss": 0.873,
1647
+ "step": 63000
1648
+ },
1649
+ {
1650
+ "epoch": 27.54880694143167,
1651
+ "grad_norm": 14412.8916015625,
1652
+ "learning_rate": 5.643339724285447e-05,
1653
+ "loss": 0.8654,
1654
+ "step": 63500
1655
+ },
1656
+ {
1657
+ "epoch": 27.765726681127983,
1658
+ "grad_norm": 11298.9375,
1659
+ "learning_rate": 5.587094827038153e-05,
1660
+ "loss": 0.8829,
1661
+ "step": 64000
1662
+ },
1663
+ {
1664
+ "epoch": 27.982646420824295,
1665
+ "grad_norm": 8732.7822265625,
1666
+ "learning_rate": 5.530661559735304e-05,
1667
+ "loss": 0.8694,
1668
+ "step": 64500
1669
+ },
1670
+ {
1671
+ "epoch": 28.0,
1672
+ "eval_loss": 0.8637914657592773,
1673
+ "eval_map": 0.4929,
1674
+ "eval_map_50": 0.8355,
1675
+ "eval_map_75": 0.5204,
1676
+ "eval_map_bus": 0.5621,
1677
+ "eval_map_car": 0.5152,
1678
+ "eval_map_container": 0.5343,
1679
+ "eval_map_large": 0.7119,
1680
+ "eval_map_medium": 0.4885,
1681
+ "eval_map_motorbike": 0.3601,
1682
+ "eval_map_small": 0.1308,
1683
+ "eval_mar_1": 0.2571,
1684
+ "eval_mar_10": 0.556,
1685
+ "eval_mar_100": 0.6029,
1686
+ "eval_mar_100_bus": 0.6851,
1687
+ "eval_mar_100_car": 0.6085,
1688
+ "eval_mar_100_container": 0.6431,
1689
+ "eval_mar_100_motorbike": 0.4749,
1690
+ "eval_mar_large": 0.7964,
1691
+ "eval_mar_medium": 0.6042,
1692
+ "eval_mar_small": 0.3162,
1693
+ "eval_runtime": 145.47,
1694
+ "eval_samples_per_second": 15.831,
1695
+ "eval_steps_per_second": 1.98,
1696
+ "step": 64540
1697
+ },
1698
+ {
1699
+ "epoch": 28.199566160520607,
1700
+ "grad_norm": 6033.21923828125,
1701
+ "learning_rate": 5.474159836931288e-05,
1702
+ "loss": 0.8805,
1703
+ "step": 65000
1704
+ },
1705
+ {
1706
+ "epoch": 28.41648590021692,
1707
+ "grad_norm": 7858.236328125,
1708
+ "learning_rate": 5.417596947365383e-05,
1709
+ "loss": 0.878,
1710
+ "step": 65500
1711
+ },
1712
+ {
1713
+ "epoch": 28.63340563991323,
1714
+ "grad_norm": 11765.7236328125,
1715
+ "learning_rate": 5.361093470095784e-05,
1716
+ "loss": 0.8746,
1717
+ "step": 66000
1718
+ },
1719
+ {
1720
+ "epoch": 28.850325379609544,
1721
+ "grad_norm": 3453.400146484375,
1722
+ "learning_rate": 5.304430229684755e-05,
1723
+ "loss": 0.8744,
1724
+ "step": 66500
1725
+ },
1726
+ {
1727
+ "epoch": 29.0,
1728
+ "eval_loss": 0.8675053119659424,
1729
+ "eval_map": 0.4883,
1730
+ "eval_map_50": 0.8358,
1731
+ "eval_map_75": 0.5122,
1732
+ "eval_map_bus": 0.5564,
1733
+ "eval_map_car": 0.5131,
1734
+ "eval_map_container": 0.5244,
1735
+ "eval_map_large": 0.6955,
1736
+ "eval_map_medium": 0.4818,
1737
+ "eval_map_motorbike": 0.3594,
1738
+ "eval_map_small": 0.1305,
1739
+ "eval_mar_1": 0.2542,
1740
+ "eval_mar_10": 0.5483,
1741
+ "eval_mar_100": 0.5948,
1742
+ "eval_mar_100_bus": 0.6693,
1743
+ "eval_mar_100_car": 0.602,
1744
+ "eval_mar_100_container": 0.6352,
1745
+ "eval_mar_100_motorbike": 0.4727,
1746
+ "eval_mar_large": 0.7802,
1747
+ "eval_mar_medium": 0.5936,
1748
+ "eval_mar_small": 0.3363,
1749
+ "eval_runtime": 145.2093,
1750
+ "eval_samples_per_second": 15.86,
1751
+ "eval_steps_per_second": 1.983,
1752
+ "step": 66845
1753
+ },
1754
+ {
1755
+ "epoch": 29.067245119305856,
1756
+ "grad_norm": 4124.80615234375,
1757
+ "learning_rate": 5.24772771768233e-05,
1758
+ "loss": 0.8763,
1759
+ "step": 67000
1760
+ },
1761
+ {
1762
+ "epoch": 29.284164859002168,
1763
+ "grad_norm": 12881.8603515625,
1764
+ "learning_rate": 5.1909932487296506e-05,
1765
+ "loss": 0.8783,
1766
+ "step": 67500
1767
+ },
1768
+ {
1769
+ "epoch": 29.501084598698483,
1770
+ "grad_norm": 11561.8046875,
1771
+ "learning_rate": 5.134234141590315e-05,
1772
+ "loss": 0.867,
1773
+ "step": 68000
1774
+ },
1775
+ {
1776
+ "epoch": 29.718004338394795,
1777
+ "grad_norm": 4195.20458984375,
1778
+ "learning_rate": 5.077571283466763e-05,
1779
+ "loss": 0.8623,
1780
+ "step": 68500
1781
+ },
1782
+ {
1783
+ "epoch": 29.934924078091107,
1784
+ "grad_norm": 6831.716796875,
1785
+ "learning_rate": 5.020784880687298e-05,
1786
+ "loss": 0.8608,
1787
+ "step": 69000
1788
+ },
1789
+ {
1790
+ "epoch": 30.0,
1791
+ "eval_loss": 0.8547199964523315,
1792
+ "eval_map": 0.5,
1793
+ "eval_map_50": 0.8401,
1794
+ "eval_map_75": 0.5304,
1795
+ "eval_map_bus": 0.5762,
1796
+ "eval_map_car": 0.5197,
1797
+ "eval_map_container": 0.5424,
1798
+ "eval_map_large": 0.7118,
1799
+ "eval_map_medium": 0.4962,
1800
+ "eval_map_motorbike": 0.3617,
1801
+ "eval_map_small": 0.1291,
1802
+ "eval_mar_1": 0.2596,
1803
+ "eval_mar_10": 0.559,
1804
+ "eval_mar_100": 0.6079,
1805
+ "eval_mar_100_bus": 0.6877,
1806
+ "eval_mar_100_car": 0.6139,
1807
+ "eval_mar_100_container": 0.6542,
1808
+ "eval_mar_100_motorbike": 0.4756,
1809
+ "eval_mar_large": 0.7898,
1810
+ "eval_mar_medium": 0.6093,
1811
+ "eval_mar_small": 0.34,
1812
+ "eval_runtime": 144.8308,
1813
+ "eval_samples_per_second": 15.901,
1814
+ "eval_steps_per_second": 1.989,
1815
+ "step": 69150
1816
+ },
1817
+ {
1818
+ "epoch": 30.15184381778742,
1819
+ "grad_norm": 9134.2763671875,
1820
+ "learning_rate": 4.9639957966519604e-05,
1821
+ "loss": 0.8683,
1822
+ "step": 69500
1823
+ },
1824
+ {
1825
+ "epoch": 30.36876355748373,
1826
+ "grad_norm": 6325.7216796875,
1827
+ "learning_rate": 4.907211357169706e-05,
1828
+ "loss": 0.8653,
1829
+ "step": 70000
1830
+ },
1831
+ {
1832
+ "epoch": 30.585683297180044,
1833
+ "grad_norm": 13625.853515625,
1834
+ "learning_rate": 4.850438887450347e-05,
1835
+ "loss": 0.8714,
1836
+ "step": 70500
1837
+ },
1838
+ {
1839
+ "epoch": 30.802603036876356,
1840
+ "grad_norm": 15019.505859375,
1841
+ "learning_rate": 4.793799193390669e-05,
1842
+ "loss": 0.8566,
1843
+ "step": 71000
1844
+ },
1845
+ {
1846
+ "epoch": 31.0,
1847
+ "eval_loss": 0.8513031005859375,
1848
+ "eval_map": 0.5,
1849
+ "eval_map_50": 0.8418,
1850
+ "eval_map_75": 0.5333,
1851
+ "eval_map_bus": 0.5714,
1852
+ "eval_map_car": 0.5239,
1853
+ "eval_map_container": 0.5422,
1854
+ "eval_map_large": 0.7031,
1855
+ "eval_map_medium": 0.4972,
1856
+ "eval_map_motorbike": 0.3626,
1857
+ "eval_map_small": 0.1282,
1858
+ "eval_mar_1": 0.2582,
1859
+ "eval_mar_10": 0.5594,
1860
+ "eval_mar_100": 0.6082,
1861
+ "eval_mar_100_bus": 0.6856,
1862
+ "eval_mar_100_car": 0.6153,
1863
+ "eval_mar_100_container": 0.6545,
1864
+ "eval_mar_100_motorbike": 0.4772,
1865
+ "eval_mar_large": 0.7875,
1866
+ "eval_mar_medium": 0.6116,
1867
+ "eval_mar_small": 0.3505,
1868
+ "eval_runtime": 144.5939,
1869
+ "eval_samples_per_second": 15.927,
1870
+ "eval_steps_per_second": 1.992,
1871
+ "step": 71455
1872
+ },
1873
+ {
1874
+ "epoch": 31.019522776572668,
1875
+ "grad_norm": 9826.4052734375,
1876
+ "learning_rate": 4.7370725711717407e-05,
1877
+ "loss": 0.8693,
1878
+ "step": 71500
1879
+ },
1880
+ {
1881
+ "epoch": 31.23644251626898,
1882
+ "grad_norm": 3330.38623046875,
1883
+ "learning_rate": 4.6803798666703856e-05,
1884
+ "loss": 0.8625,
1885
+ "step": 72000
1886
+ },
1887
+ {
1888
+ "epoch": 31.453362255965292,
1889
+ "grad_norm": 4638.60107421875,
1890
+ "learning_rate": 4.623728393262576e-05,
1891
+ "loss": 0.8672,
1892
+ "step": 72500
1893
+ },
1894
+ {
1895
+ "epoch": 31.670281995661604,
1896
+ "grad_norm": 10907.923828125,
1897
+ "learning_rate": 4.5672386115774904e-05,
1898
+ "loss": 0.8625,
1899
+ "step": 73000
1900
+ },
1901
+ {
1902
+ "epoch": 31.887201735357916,
1903
+ "grad_norm": 3609.915283203125,
1904
+ "learning_rate": 4.5106913993021883e-05,
1905
+ "loss": 0.8586,
1906
+ "step": 73500
1907
+ },
1908
+ {
1909
+ "epoch": 32.0,
1910
+ "eval_loss": 0.8532082438468933,
1911
+ "eval_map": 0.5035,
1912
+ "eval_map_50": 0.8414,
1913
+ "eval_map_75": 0.538,
1914
+ "eval_map_bus": 0.5799,
1915
+ "eval_map_car": 0.5235,
1916
+ "eval_map_container": 0.5481,
1917
+ "eval_map_large": 0.7084,
1918
+ "eval_map_medium": 0.4993,
1919
+ "eval_map_motorbike": 0.3626,
1920
+ "eval_map_small": 0.1292,
1921
+ "eval_mar_1": 0.2603,
1922
+ "eval_mar_10": 0.5621,
1923
+ "eval_mar_100": 0.6094,
1924
+ "eval_mar_100_bus": 0.6893,
1925
+ "eval_mar_100_car": 0.6139,
1926
+ "eval_mar_100_container": 0.6587,
1927
+ "eval_mar_100_motorbike": 0.4758,
1928
+ "eval_mar_large": 0.7924,
1929
+ "eval_mar_medium": 0.6135,
1930
+ "eval_mar_small": 0.3286,
1931
+ "eval_runtime": 145.9881,
1932
+ "eval_samples_per_second": 15.775,
1933
+ "eval_steps_per_second": 1.973,
1934
+ "step": 73760
1935
+ },
1936
+ {
1937
+ "epoch": 32.10412147505423,
1938
+ "grad_norm": 11651.8828125,
1939
+ "learning_rate": 4.4542073079838684e-05,
1940
+ "loss": 0.8618,
1941
+ "step": 74000
1942
+ },
1943
+ {
1944
+ "epoch": 32.321041214750544,
1945
+ "grad_norm": 17436.388671875,
1946
+ "learning_rate": 4.39779362408734e-05,
1947
+ "loss": 0.8542,
1948
+ "step": 74500
1949
+ },
1950
+ {
1951
+ "epoch": 32.53796095444685,
1952
+ "grad_norm": 20459.701171875,
1953
+ "learning_rate": 4.341457624994828e-05,
1954
+ "loss": 0.8554,
1955
+ "step": 75000
1956
+ },
1957
+ {
1958
+ "epoch": 32.75488069414317,
1959
+ "grad_norm": 5393.4375,
1960
+ "learning_rate": 4.285318990553778e-05,
1961
+ "loss": 0.8528,
1962
+ "step": 75500
1963
+ },
1964
+ {
1965
+ "epoch": 32.971800433839476,
1966
+ "grad_norm": 4986.16796875,
1967
+ "learning_rate": 4.229159960541767e-05,
1968
+ "loss": 0.8598,
1969
+ "step": 76000
1970
+ },
1971
+ {
1972
+ "epoch": 33.0,
1973
+ "eval_loss": 0.8493744134902954,
1974
+ "eval_map": 0.499,
1975
+ "eval_map_50": 0.8415,
1976
+ "eval_map_75": 0.5323,
1977
+ "eval_map_bus": 0.5708,
1978
+ "eval_map_car": 0.5215,
1979
+ "eval_map_container": 0.5432,
1980
+ "eval_map_large": 0.7104,
1981
+ "eval_map_medium": 0.4969,
1982
+ "eval_map_motorbike": 0.3606,
1983
+ "eval_map_small": 0.1301,
1984
+ "eval_mar_1": 0.2582,
1985
+ "eval_mar_10": 0.5605,
1986
+ "eval_mar_100": 0.6084,
1987
+ "eval_mar_100_bus": 0.6857,
1988
+ "eval_mar_100_car": 0.615,
1989
+ "eval_mar_100_container": 0.6551,
1990
+ "eval_mar_100_motorbike": 0.4778,
1991
+ "eval_mar_large": 0.7942,
1992
+ "eval_mar_medium": 0.6113,
1993
+ "eval_mar_small": 0.3277,
1994
+ "eval_runtime": 144.7755,
1995
+ "eval_samples_per_second": 15.907,
1996
+ "eval_steps_per_second": 1.989,
1997
+ "step": 76065
1998
+ },
1999
+ {
2000
+ "epoch": 33.18872017353579,
2001
+ "grad_norm": 6646.677734375,
2002
+ "learning_rate": 4.173212384261402e-05,
2003
+ "loss": 0.8671,
2004
+ "step": 76500
2005
+ },
2006
+ {
2007
+ "epoch": 33.40563991323211,
2008
+ "grad_norm": 4728.1552734375,
2009
+ "learning_rate": 4.1172592425996225e-05,
2010
+ "loss": 0.8671,
2011
+ "step": 77000
2012
+ },
2013
+ {
2014
+ "epoch": 33.622559652928416,
2015
+ "grad_norm": 9825.3720703125,
2016
+ "learning_rate": 4.0614199747616364e-05,
2017
+ "loss": 0.8668,
2018
+ "step": 77500
2019
+ },
2020
+ {
2021
+ "epoch": 33.83947939262473,
2022
+ "grad_norm": 3324.343505859375,
2023
+ "learning_rate": 4.0057017840298235e-05,
2024
+ "loss": 0.8728,
2025
+ "step": 78000
2026
+ },
2027
+ {
2028
+ "epoch": 34.0,
2029
+ "eval_loss": 0.8495970368385315,
2030
+ "eval_map": 0.5027,
2031
+ "eval_map_50": 0.841,
2032
+ "eval_map_75": 0.5357,
2033
+ "eval_map_bus": 0.5809,
2034
+ "eval_map_car": 0.5196,
2035
+ "eval_map_container": 0.5452,
2036
+ "eval_map_large": 0.715,
2037
+ "eval_map_medium": 0.4969,
2038
+ "eval_map_motorbike": 0.3652,
2039
+ "eval_map_small": 0.1332,
2040
+ "eval_mar_1": 0.2609,
2041
+ "eval_mar_10": 0.5617,
2042
+ "eval_mar_100": 0.6105,
2043
+ "eval_mar_100_bus": 0.6941,
2044
+ "eval_mar_100_car": 0.6116,
2045
+ "eval_mar_100_container": 0.6565,
2046
+ "eval_mar_100_motorbike": 0.4799,
2047
+ "eval_mar_large": 0.7994,
2048
+ "eval_mar_medium": 0.6147,
2049
+ "eval_mar_small": 0.3335,
2050
+ "eval_runtime": 145.9053,
2051
+ "eval_samples_per_second": 15.784,
2052
+ "eval_steps_per_second": 1.974,
2053
+ "step": 78370
2054
+ },
2055
+ {
2056
+ "epoch": 34.05639913232104,
2057
+ "grad_norm": 20089.041015625,
2058
+ "learning_rate": 3.950111858067586e-05,
2059
+ "loss": 0.8606,
2060
+ "step": 78500
2061
+ },
2062
+ {
2063
+ "epoch": 34.273318872017356,
2064
+ "grad_norm": 22944.431640625,
2065
+ "learning_rate": 3.894657367992129e-05,
2066
+ "loss": 0.8592,
2067
+ "step": 79000
2068
+ },
2069
+ {
2070
+ "epoch": 34.490238611713664,
2071
+ "grad_norm": 6247.44970703125,
2072
+ "learning_rate": 3.8393454674493866e-05,
2073
+ "loss": 0.8535,
2074
+ "step": 79500
2075
+ },
2076
+ {
2077
+ "epoch": 34.70715835140998,
2078
+ "grad_norm": 8178.81982421875,
2079
+ "learning_rate": 3.7841832916912035e-05,
2080
+ "loss": 0.8499,
2081
+ "step": 80000
2082
+ },
2083
+ {
2084
+ "epoch": 34.92407809110629,
2085
+ "grad_norm": 6138.80712890625,
2086
+ "learning_rate": 3.7291779566548805e-05,
2087
+ "loss": 0.8496,
2088
+ "step": 80500
2089
+ },
2090
+ {
2091
+ "epoch": 35.0,
2092
+ "eval_loss": 0.8385424613952637,
2093
+ "eval_map": 0.5081,
2094
+ "eval_map_50": 0.8476,
2095
+ "eval_map_75": 0.5424,
2096
+ "eval_map_bus": 0.5828,
2097
+ "eval_map_car": 0.5317,
2098
+ "eval_map_container": 0.5475,
2099
+ "eval_map_large": 0.7206,
2100
+ "eval_map_medium": 0.504,
2101
+ "eval_map_motorbike": 0.3703,
2102
+ "eval_map_small": 0.1423,
2103
+ "eval_mar_1": 0.2623,
2104
+ "eval_mar_10": 0.5665,
2105
+ "eval_mar_100": 0.6148,
2106
+ "eval_mar_100_bus": 0.6954,
2107
+ "eval_mar_100_car": 0.6213,
2108
+ "eval_mar_100_container": 0.6591,
2109
+ "eval_mar_100_motorbike": 0.4833,
2110
+ "eval_mar_large": 0.8048,
2111
+ "eval_mar_medium": 0.6188,
2112
+ "eval_mar_small": 0.3475,
2113
+ "eval_runtime": 144.7118,
2114
+ "eval_samples_per_second": 15.914,
2115
+ "eval_steps_per_second": 1.99,
2116
+ "step": 80675
2117
+ }
2118
+ ],
2119
+ "logging_steps": 500,
2120
+ "max_steps": 138300,
2121
+ "num_input_tokens_seen": 0,
2122
+ "num_train_epochs": 60,
2123
+ "save_steps": 500,
2124
+ "stateful_callbacks": {
2125
+ "TrainerControl": {
2126
+ "args": {
2127
+ "should_epoch_stop": false,
2128
+ "should_evaluate": false,
2129
+ "should_log": false,
2130
+ "should_save": true,
2131
+ "should_training_stop": false
2132
+ },
2133
+ "attributes": {}
2134
+ }
2135
+ },
2136
+ "total_flos": 1.4836567123062817e+20,
2137
+ "train_batch_size": 4,
2138
+ "trial_name": null,
2139
+ "trial_params": null
2140
+ }
checkpoint-80675/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e79ce30220cc0f492cce991ef1b82af549d14310ccde1c3ab2fa0987f62e3936
3
+ size 5240
checkpoint-82980/config.json ADDED
@@ -0,0 +1,67 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "datamonster/ms-cond-detr-res-50-vehicles",
3
+ "activation_dropout": 0.0,
4
+ "activation_function": "relu",
5
+ "architectures": [
6
+ "ConditionalDetrForObjectDetection"
7
+ ],
8
+ "attention_dropout": 0.0,
9
+ "auxiliary_loss": false,
10
+ "backbone": "resnet50",
11
+ "backbone_config": null,
12
+ "backbone_kwargs": {
13
+ "in_chans": 3,
14
+ "out_indices": [
15
+ 1,
16
+ 2,
17
+ 3,
18
+ 4
19
+ ]
20
+ },
21
+ "bbox_cost": 5,
22
+ "bbox_loss_coefficient": 5,
23
+ "class_cost": 2,
24
+ "cls_loss_coefficient": 2,
25
+ "d_model": 256,
26
+ "decoder_attention_heads": 8,
27
+ "decoder_ffn_dim": 2048,
28
+ "decoder_layerdrop": 0.0,
29
+ "decoder_layers": 6,
30
+ "dice_loss_coefficient": 1,
31
+ "dilation": false,
32
+ "dropout": 0.1,
33
+ "encoder_attention_heads": 8,
34
+ "encoder_ffn_dim": 2048,
35
+ "encoder_layerdrop": 0.0,
36
+ "encoder_layers": 6,
37
+ "focal_alpha": 0.25,
38
+ "giou_cost": 2,
39
+ "giou_loss_coefficient": 2,
40
+ "id2label": {
41
+ "0": "motorbike",
42
+ "1": "car",
43
+ "2": "bus",
44
+ "3": "container"
45
+ },
46
+ "init_std": 0.02,
47
+ "init_xavier_std": 1.0,
48
+ "is_encoder_decoder": true,
49
+ "label2id": {
50
+ "bus": 2,
51
+ "car": 1,
52
+ "container": 3,
53
+ "motorbike": 0
54
+ },
55
+ "mask_loss_coefficient": 1,
56
+ "max_position_embeddings": 1024,
57
+ "model_type": "conditional_detr",
58
+ "num_channels": 3,
59
+ "num_hidden_layers": 6,
60
+ "num_queries": 300,
61
+ "position_embedding_type": "sine",
62
+ "scale_embedding": false,
63
+ "torch_dtype": "float32",
64
+ "transformers_version": "4.45.1",
65
+ "use_pretrained_backbone": true,
66
+ "use_timm_backbone": true
67
+ }
checkpoint-82980/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:19e8ccc78453da24cc2bf7f444ceca090c700aa3192a6ece9f161354d623c765
3
+ size 174078768
checkpoint-82980/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:303e4f468002a7f6069aa3738be031c628a94d0cb1692e75fa13e8bb645d1cf5
3
+ size 187563450
checkpoint-82980/preprocessor_config.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_convert_annotations": true,
3
+ "do_normalize": true,
4
+ "do_pad": true,
5
+ "do_rescale": true,
6
+ "do_resize": true,
7
+ "format": "coco_detection",
8
+ "image_mean": [
9
+ 0.485,
10
+ 0.456,
11
+ 0.406
12
+ ],
13
+ "image_processor_type": "ConditionalDetrImageProcessor",
14
+ "image_std": [
15
+ 0.229,
16
+ 0.224,
17
+ 0.225
18
+ ],
19
+ "pad_size": {
20
+ "height": 768,
21
+ "width": 768
22
+ },
23
+ "resample": 2,
24
+ "rescale_factor": 0.00392156862745098,
25
+ "size": {
26
+ "max_height": 768,
27
+ "max_width": 768
28
+ }
29
+ }
checkpoint-82980/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:16025297b40c1b9324f8606428cf9ba1132c0408b13420431cb554068d34e21d
3
+ size 14244
checkpoint-82980/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6fa5562b32f18d5471ebde99c8d014cd91d547d6c8005ac58c7eccd0cbba6727
3
+ size 1000
checkpoint-82980/trainer_state.json ADDED
@@ -0,0 +1,2196 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.5081,
3
+ "best_model_checkpoint": "ms-cond-detr-res-50-vehicles/checkpoint-80675",
4
+ "epoch": 36.0,
5
+ "eval_steps": 500,
6
+ "global_step": 82980,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.21691973969631237,
13
+ "grad_norm": 38753.67578125,
14
+ "learning_rate": 9.999686465963656e-05,
15
+ "loss": 1.2341,
16
+ "step": 500
17
+ },
18
+ {
19
+ "epoch": 0.43383947939262474,
20
+ "grad_norm": 14156.765625,
21
+ "learning_rate": 9.998728034076959e-05,
22
+ "loss": 1.1441,
23
+ "step": 1000
24
+ },
25
+ {
26
+ "epoch": 0.6507592190889371,
27
+ "grad_norm": 7688.85986328125,
28
+ "learning_rate": 9.99712861472983e-05,
29
+ "loss": 1.1357,
30
+ "step": 1500
31
+ },
32
+ {
33
+ "epoch": 0.8676789587852495,
34
+ "grad_norm": 10963.5341796875,
35
+ "learning_rate": 9.994882003818517e-05,
36
+ "loss": 1.1322,
37
+ "step": 2000
38
+ },
39
+ {
40
+ "epoch": 1.0,
41
+ "eval_loss": 1.0705492496490479,
42
+ "eval_map": 0.3886,
43
+ "eval_map_50": 0.7276,
44
+ "eval_map_75": 0.371,
45
+ "eval_map_bus": 0.4553,
46
+ "eval_map_car": 0.4267,
47
+ "eval_map_container": 0.4051,
48
+ "eval_map_large": 0.6092,
49
+ "eval_map_medium": 0.3725,
50
+ "eval_map_motorbike": 0.2674,
51
+ "eval_map_small": 0.0755,
52
+ "eval_mar_1": 0.2228,
53
+ "eval_mar_10": 0.4736,
54
+ "eval_mar_100": 0.5226,
55
+ "eval_mar_100_bus": 0.6014,
56
+ "eval_mar_100_car": 0.5398,
57
+ "eval_mar_100_container": 0.5541,
58
+ "eval_mar_100_motorbike": 0.395,
59
+ "eval_mar_large": 0.7004,
60
+ "eval_mar_medium": 0.5176,
61
+ "eval_mar_small": 0.2475,
62
+ "eval_runtime": 149.8641,
63
+ "eval_samples_per_second": 15.367,
64
+ "eval_steps_per_second": 1.922,
65
+ "step": 2305
66
+ },
67
+ {
68
+ "epoch": 1.0845986984815619,
69
+ "grad_norm": 8778.8525390625,
70
+ "learning_rate": 9.991991051623087e-05,
71
+ "loss": 1.1204,
72
+ "step": 2500
73
+ },
74
+ {
75
+ "epoch": 1.3015184381778742,
76
+ "grad_norm": 8215.1748046875,
77
+ "learning_rate": 9.988456131077249e-05,
78
+ "loss": 1.0925,
79
+ "step": 3000
80
+ },
81
+ {
82
+ "epoch": 1.5184381778741867,
83
+ "grad_norm": 15722.8212890625,
84
+ "learning_rate": 9.984277698186814e-05,
85
+ "loss": 1.0867,
86
+ "step": 3500
87
+ },
88
+ {
89
+ "epoch": 1.735357917570499,
90
+ "grad_norm": 17670.681640625,
91
+ "learning_rate": 9.979456291970887e-05,
92
+ "loss": 1.0714,
93
+ "step": 4000
94
+ },
95
+ {
96
+ "epoch": 1.9522776572668112,
97
+ "grad_norm": 15503.669921875,
98
+ "learning_rate": 9.974004102519452e-05,
99
+ "loss": 1.0605,
100
+ "step": 4500
101
+ },
102
+ {
103
+ "epoch": 2.0,
104
+ "eval_loss": 1.0418221950531006,
105
+ "eval_map": 0.3977,
106
+ "eval_map_50": 0.7371,
107
+ "eval_map_75": 0.389,
108
+ "eval_map_bus": 0.4529,
109
+ "eval_map_car": 0.4318,
110
+ "eval_map_container": 0.4316,
111
+ "eval_map_large": 0.6212,
112
+ "eval_map_medium": 0.3842,
113
+ "eval_map_motorbike": 0.2745,
114
+ "eval_map_small": 0.0841,
115
+ "eval_mar_1": 0.2265,
116
+ "eval_mar_10": 0.4923,
117
+ "eval_mar_100": 0.5393,
118
+ "eval_mar_100_bus": 0.6241,
119
+ "eval_mar_100_car": 0.5529,
120
+ "eval_mar_100_container": 0.5848,
121
+ "eval_mar_100_motorbike": 0.3954,
122
+ "eval_mar_large": 0.7101,
123
+ "eval_mar_medium": 0.5389,
124
+ "eval_mar_small": 0.2511,
125
+ "eval_runtime": 146.5587,
126
+ "eval_samples_per_second": 15.714,
127
+ "eval_steps_per_second": 1.965,
128
+ "step": 4610
129
+ },
130
+ {
131
+ "epoch": 2.1691973969631237,
132
+ "grad_norm": 3109.66162109375,
133
+ "learning_rate": 9.967899980939251e-05,
134
+ "loss": 1.0591,
135
+ "step": 5000
136
+ },
137
+ {
138
+ "epoch": 2.3861171366594363,
139
+ "grad_norm": 5812.068359375,
140
+ "learning_rate": 9.961154998764017e-05,
141
+ "loss": 1.0676,
142
+ "step": 5500
143
+ },
144
+ {
145
+ "epoch": 2.6030368763557483,
146
+ "grad_norm": 7711.87158203125,
147
+ "learning_rate": 9.953770026098485e-05,
148
+ "loss": 1.0673,
149
+ "step": 6000
150
+ },
151
+ {
152
+ "epoch": 2.819956616052061,
153
+ "grad_norm": 11271.8916015625,
154
+ "learning_rate": 9.945746015606358e-05,
155
+ "loss": 1.0634,
156
+ "step": 6500
157
+ },
158
+ {
159
+ "epoch": 3.0,
160
+ "eval_loss": 1.0088645219802856,
161
+ "eval_map": 0.4231,
162
+ "eval_map_50": 0.758,
163
+ "eval_map_75": 0.4223,
164
+ "eval_map_bus": 0.4788,
165
+ "eval_map_car": 0.4565,
166
+ "eval_map_container": 0.4648,
167
+ "eval_map_large": 0.658,
168
+ "eval_map_medium": 0.4084,
169
+ "eval_map_motorbike": 0.2923,
170
+ "eval_map_small": 0.09,
171
+ "eval_mar_1": 0.2352,
172
+ "eval_mar_10": 0.5048,
173
+ "eval_mar_100": 0.555,
174
+ "eval_mar_100_bus": 0.6323,
175
+ "eval_mar_100_car": 0.5672,
176
+ "eval_mar_100_container": 0.6028,
177
+ "eval_mar_100_motorbike": 0.4178,
178
+ "eval_mar_large": 0.7546,
179
+ "eval_mar_medium": 0.5521,
180
+ "eval_mar_small": 0.2802,
181
+ "eval_runtime": 147.4693,
182
+ "eval_samples_per_second": 15.617,
183
+ "eval_steps_per_second": 1.953,
184
+ "step": 6915
185
+ },
186
+ {
187
+ "epoch": 3.036876355748373,
188
+ "grad_norm": 20180.71484375,
189
+ "learning_rate": 9.937084002387407e-05,
190
+ "loss": 1.0418,
191
+ "step": 7000
192
+ },
193
+ {
194
+ "epoch": 3.2537960954446854,
195
+ "grad_norm": 11818.376953125,
196
+ "learning_rate": 9.927804336468942e-05,
197
+ "loss": 1.0386,
198
+ "step": 7500
199
+ },
200
+ {
201
+ "epoch": 3.470715835140998,
202
+ "grad_norm": 9070.5185546875,
203
+ "learning_rate": 9.917871022281522e-05,
204
+ "loss": 1.0635,
205
+ "step": 8000
206
+ },
207
+ {
208
+ "epoch": 3.6876355748373104,
209
+ "grad_norm": 7446.32666015625,
210
+ "learning_rate": 9.907303301249819e-05,
211
+ "loss": 1.0466,
212
+ "step": 8500
213
+ },
214
+ {
215
+ "epoch": 3.9045553145336225,
216
+ "grad_norm": 2428.818603515625,
217
+ "learning_rate": 9.896102536613032e-05,
218
+ "loss": 1.0288,
219
+ "step": 9000
220
+ },
221
+ {
222
+ "epoch": 4.0,
223
+ "eval_loss": 1.002821922302246,
224
+ "eval_map": 0.4297,
225
+ "eval_map_50": 0.7647,
226
+ "eval_map_75": 0.429,
227
+ "eval_map_bus": 0.4925,
228
+ "eval_map_car": 0.4582,
229
+ "eval_map_container": 0.4685,
230
+ "eval_map_large": 0.6667,
231
+ "eval_map_medium": 0.4184,
232
+ "eval_map_motorbike": 0.2997,
233
+ "eval_map_small": 0.0816,
234
+ "eval_mar_1": 0.2383,
235
+ "eval_mar_10": 0.5069,
236
+ "eval_mar_100": 0.5549,
237
+ "eval_mar_100_bus": 0.634,
238
+ "eval_mar_100_car": 0.5649,
239
+ "eval_mar_100_container": 0.5965,
240
+ "eval_mar_100_motorbike": 0.4243,
241
+ "eval_mar_large": 0.7579,
242
+ "eval_mar_medium": 0.5516,
243
+ "eval_mar_small": 0.2596,
244
+ "eval_runtime": 146.3143,
245
+ "eval_samples_per_second": 15.74,
246
+ "eval_steps_per_second": 1.968,
247
+ "step": 9220
248
+ },
249
+ {
250
+ "epoch": 4.1214750542299345,
251
+ "grad_norm": 4431.853515625,
252
+ "learning_rate": 9.88429446733437e-05,
253
+ "loss": 1.0289,
254
+ "step": 9500
255
+ },
256
+ {
257
+ "epoch": 4.3383947939262475,
258
+ "grad_norm": 16948.7890625,
259
+ "learning_rate": 9.871833290237144e-05,
260
+ "loss": 1.031,
261
+ "step": 10000
262
+ },
263
+ {
264
+ "epoch": 4.55531453362256,
265
+ "grad_norm": 5605.24951171875,
266
+ "learning_rate": 9.858743645176947e-05,
267
+ "loss": 1.0303,
268
+ "step": 10500
269
+ },
270
+ {
271
+ "epoch": 4.7722342733188725,
272
+ "grad_norm": 16235.091796875,
273
+ "learning_rate": 9.845027220721935e-05,
274
+ "loss": 1.0206,
275
+ "step": 11000
276
+ },
277
+ {
278
+ "epoch": 4.989154013015185,
279
+ "grad_norm": 20350.36328125,
280
+ "learning_rate": 9.830685786294999e-05,
281
+ "loss": 1.0226,
282
+ "step": 11500
283
+ },
284
+ {
285
+ "epoch": 5.0,
286
+ "eval_loss": 0.9774508476257324,
287
+ "eval_map": 0.4366,
288
+ "eval_map_50": 0.7785,
289
+ "eval_map_75": 0.4388,
290
+ "eval_map_bus": 0.5023,
291
+ "eval_map_car": 0.4658,
292
+ "eval_map_container": 0.4766,
293
+ "eval_map_large": 0.6605,
294
+ "eval_map_medium": 0.4243,
295
+ "eval_map_motorbike": 0.3015,
296
+ "eval_map_small": 0.0942,
297
+ "eval_mar_1": 0.2402,
298
+ "eval_mar_10": 0.5173,
299
+ "eval_mar_100": 0.5666,
300
+ "eval_mar_100_bus": 0.6534,
301
+ "eval_mar_100_car": 0.5747,
302
+ "eval_mar_100_container": 0.6145,
303
+ "eval_mar_100_motorbike": 0.4239,
304
+ "eval_mar_large": 0.751,
305
+ "eval_mar_medium": 0.5659,
306
+ "eval_mar_small": 0.2757,
307
+ "eval_runtime": 146.1171,
308
+ "eval_samples_per_second": 15.761,
309
+ "eval_steps_per_second": 1.971,
310
+ "step": 11525
311
+ },
312
+ {
313
+ "epoch": 5.206073752711497,
314
+ "grad_norm": 11680.388671875,
315
+ "learning_rate": 9.815721191945502e-05,
316
+ "loss": 1.0021,
317
+ "step": 12000
318
+ },
319
+ {
320
+ "epoch": 5.422993492407809,
321
+ "grad_norm": 6468.74072265625,
322
+ "learning_rate": 9.800167158422275e-05,
323
+ "loss": 1.0049,
324
+ "step": 12500
325
+ },
326
+ {
327
+ "epoch": 5.639913232104122,
328
+ "grad_norm": 10755.380859375,
329
+ "learning_rate": 9.78396335205612e-05,
330
+ "loss": 1.016,
331
+ "step": 13000
332
+ },
333
+ {
334
+ "epoch": 5.856832971800434,
335
+ "grad_norm": 6588.42578125,
336
+ "learning_rate": 9.767142412975505e-05,
337
+ "loss": 1.022,
338
+ "step": 13500
339
+ },
340
+ {
341
+ "epoch": 6.0,
342
+ "eval_loss": 0.9850358366966248,
343
+ "eval_map": 0.4356,
344
+ "eval_map_50": 0.7775,
345
+ "eval_map_75": 0.4345,
346
+ "eval_map_bus": 0.5032,
347
+ "eval_map_car": 0.455,
348
+ "eval_map_container": 0.4783,
349
+ "eval_map_large": 0.661,
350
+ "eval_map_medium": 0.4213,
351
+ "eval_map_motorbike": 0.3059,
352
+ "eval_map_small": 0.106,
353
+ "eval_mar_1": 0.2402,
354
+ "eval_mar_10": 0.514,
355
+ "eval_mar_100": 0.563,
356
+ "eval_mar_100_bus": 0.651,
357
+ "eval_mar_100_car": 0.5596,
358
+ "eval_mar_100_container": 0.6163,
359
+ "eval_mar_100_motorbike": 0.4251,
360
+ "eval_mar_large": 0.7479,
361
+ "eval_mar_medium": 0.5634,
362
+ "eval_mar_small": 0.2541,
363
+ "eval_runtime": 145.7736,
364
+ "eval_samples_per_second": 15.798,
365
+ "eval_steps_per_second": 1.976,
366
+ "step": 13830
367
+ },
368
+ {
369
+ "epoch": 6.073752711496746,
370
+ "grad_norm": 8365.4345703125,
371
+ "learning_rate": 9.749706511086642e-05,
372
+ "loss": 1.0065,
373
+ "step": 14000
374
+ },
375
+ {
376
+ "epoch": 6.290672451193059,
377
+ "grad_norm": 7092.89306640625,
378
+ "learning_rate": 9.731694602810796e-05,
379
+ "loss": 1.0004,
380
+ "step": 14500
381
+ },
382
+ {
383
+ "epoch": 6.507592190889371,
384
+ "grad_norm": 7779.22314453125,
385
+ "learning_rate": 9.713036820450053e-05,
386
+ "loss": 1.0193,
387
+ "step": 15000
388
+ },
389
+ {
390
+ "epoch": 6.724511930585683,
391
+ "grad_norm": 5161.82958984375,
392
+ "learning_rate": 9.693771054918806e-05,
393
+ "loss": 0.9992,
394
+ "step": 15500
395
+ },
396
+ {
397
+ "epoch": 6.941431670281996,
398
+ "grad_norm": 5342.78662109375,
399
+ "learning_rate": 9.673899791506617e-05,
400
+ "loss": 0.9936,
401
+ "step": 16000
402
+ },
403
+ {
404
+ "epoch": 7.0,
405
+ "eval_loss": 0.9677829742431641,
406
+ "eval_map": 0.4394,
407
+ "eval_map_50": 0.781,
408
+ "eval_map_75": 0.4415,
409
+ "eval_map_bus": 0.5059,
410
+ "eval_map_car": 0.4716,
411
+ "eval_map_container": 0.4731,
412
+ "eval_map_large": 0.6636,
413
+ "eval_map_medium": 0.4308,
414
+ "eval_map_motorbike": 0.307,
415
+ "eval_map_small": 0.109,
416
+ "eval_mar_1": 0.2401,
417
+ "eval_mar_10": 0.5184,
418
+ "eval_mar_100": 0.5674,
419
+ "eval_mar_100_bus": 0.6546,
420
+ "eval_mar_100_car": 0.5736,
421
+ "eval_mar_100_container": 0.6126,
422
+ "eval_mar_100_motorbike": 0.4289,
423
+ "eval_mar_large": 0.7468,
424
+ "eval_mar_medium": 0.5696,
425
+ "eval_mar_small": 0.2775,
426
+ "eval_runtime": 147.1057,
427
+ "eval_samples_per_second": 15.655,
428
+ "eval_steps_per_second": 1.958,
429
+ "step": 16135
430
+ },
431
+ {
432
+ "epoch": 7.158351409978308,
433
+ "grad_norm": 17945.982421875,
434
+ "learning_rate": 9.653425593612456e-05,
435
+ "loss": 0.9817,
436
+ "step": 16500
437
+ },
438
+ {
439
+ "epoch": 7.375271149674621,
440
+ "grad_norm": 2631.81591796875,
441
+ "learning_rate": 9.632393848695017e-05,
442
+ "loss": 0.9845,
443
+ "step": 17000
444
+ },
445
+ {
446
+ "epoch": 7.592190889370933,
447
+ "grad_norm": 1825.7994384765625,
448
+ "learning_rate": 9.61072297519293e-05,
449
+ "loss": 0.9839,
450
+ "step": 17500
451
+ },
452
+ {
453
+ "epoch": 7.809110629067245,
454
+ "grad_norm": 8929.837890625,
455
+ "learning_rate": 9.588457317037219e-05,
456
+ "loss": 0.9908,
457
+ "step": 18000
458
+ },
459
+ {
460
+ "epoch": 8.0,
461
+ "eval_loss": 0.9544239640235901,
462
+ "eval_map": 0.4515,
463
+ "eval_map_50": 0.7929,
464
+ "eval_map_75": 0.459,
465
+ "eval_map_bus": 0.5215,
466
+ "eval_map_car": 0.4759,
467
+ "eval_map_container": 0.4956,
468
+ "eval_map_large": 0.6707,
469
+ "eval_map_medium": 0.4407,
470
+ "eval_map_motorbike": 0.3131,
471
+ "eval_map_small": 0.103,
472
+ "eval_mar_1": 0.2439,
473
+ "eval_mar_10": 0.5243,
474
+ "eval_mar_100": 0.573,
475
+ "eval_mar_100_bus": 0.6558,
476
+ "eval_mar_100_car": 0.579,
477
+ "eval_mar_100_container": 0.6217,
478
+ "eval_mar_100_motorbike": 0.4355,
479
+ "eval_mar_large": 0.7575,
480
+ "eval_mar_medium": 0.5708,
481
+ "eval_mar_small": 0.3107,
482
+ "eval_runtime": 147.1314,
483
+ "eval_samples_per_second": 15.653,
484
+ "eval_steps_per_second": 1.957,
485
+ "step": 18440
486
+ },
487
+ {
488
+ "epoch": 8.026030368763557,
489
+ "grad_norm": 9851.580078125,
490
+ "learning_rate": 9.565599746504495e-05,
491
+ "loss": 0.9914,
492
+ "step": 18500
493
+ },
494
+ {
495
+ "epoch": 8.242950108459869,
496
+ "grad_norm": 46131.0703125,
497
+ "learning_rate": 9.542153212228251e-05,
498
+ "loss": 0.9789,
499
+ "step": 19000
500
+ },
501
+ {
502
+ "epoch": 8.459869848156183,
503
+ "grad_norm": 40477.8671875,
504
+ "learning_rate": 9.518120738818473e-05,
505
+ "loss": 0.9642,
506
+ "step": 19500
507
+ },
508
+ {
509
+ "epoch": 8.676789587852495,
510
+ "grad_norm": 21795.466796875,
511
+ "learning_rate": 9.493505426471479e-05,
512
+ "loss": 0.9637,
513
+ "step": 20000
514
+ },
515
+ {
516
+ "epoch": 8.893709327548807,
517
+ "grad_norm": 5294.40673828125,
518
+ "learning_rate": 9.468361416878137e-05,
519
+ "loss": 0.9853,
520
+ "step": 20500
521
+ },
522
+ {
523
+ "epoch": 9.0,
524
+ "eval_loss": 0.9467804431915283,
525
+ "eval_map": 0.4529,
526
+ "eval_map_50": 0.7951,
527
+ "eval_map_75": 0.46,
528
+ "eval_map_bus": 0.5273,
529
+ "eval_map_car": 0.4728,
530
+ "eval_map_container": 0.4969,
531
+ "eval_map_large": 0.6735,
532
+ "eval_map_medium": 0.4421,
533
+ "eval_map_motorbike": 0.3144,
534
+ "eval_map_small": 0.1016,
535
+ "eval_mar_1": 0.2466,
536
+ "eval_mar_10": 0.5271,
537
+ "eval_mar_100": 0.5772,
538
+ "eval_mar_100_bus": 0.6592,
539
+ "eval_mar_100_car": 0.5831,
540
+ "eval_mar_100_container": 0.6281,
541
+ "eval_mar_100_motorbike": 0.4385,
542
+ "eval_mar_large": 0.7607,
543
+ "eval_mar_medium": 0.5767,
544
+ "eval_mar_small": 0.2947,
545
+ "eval_runtime": 146.7224,
546
+ "eval_samples_per_second": 15.696,
547
+ "eval_steps_per_second": 1.963,
548
+ "step": 20745
549
+ },
550
+ {
551
+ "epoch": 9.11062906724512,
552
+ "grad_norm": 20100.53515625,
553
+ "learning_rate": 9.442591177115253e-05,
554
+ "loss": 0.973,
555
+ "step": 21000
556
+ },
557
+ {
558
+ "epoch": 9.327548806941431,
559
+ "grad_norm": 4432.14404296875,
560
+ "learning_rate": 9.416247841751377e-05,
561
+ "loss": 0.9844,
562
+ "step": 21500
563
+ },
564
+ {
565
+ "epoch": 9.544468546637743,
566
+ "grad_norm": 5448.53369140625,
567
+ "learning_rate": 9.389334809084711e-05,
568
+ "loss": 0.9783,
569
+ "step": 22000
570
+ },
571
+ {
572
+ "epoch": 9.761388286334057,
573
+ "grad_norm": 18744.869140625,
574
+ "learning_rate": 9.361855550904582e-05,
575
+ "loss": 0.9591,
576
+ "step": 22500
577
+ },
578
+ {
579
+ "epoch": 9.97830802603037,
580
+ "grad_norm": 9256.87890625,
581
+ "learning_rate": 9.333870255084278e-05,
582
+ "loss": 0.9536,
583
+ "step": 23000
584
+ },
585
+ {
586
+ "epoch": 10.0,
587
+ "eval_loss": 0.9346117377281189,
588
+ "eval_map": 0.4603,
589
+ "eval_map_50": 0.8028,
590
+ "eval_map_75": 0.4763,
591
+ "eval_map_bus": 0.5317,
592
+ "eval_map_car": 0.4808,
593
+ "eval_map_container": 0.5064,
594
+ "eval_map_large": 0.6826,
595
+ "eval_map_medium": 0.4525,
596
+ "eval_map_motorbike": 0.3221,
597
+ "eval_map_small": 0.108,
598
+ "eval_mar_1": 0.2479,
599
+ "eval_mar_10": 0.532,
600
+ "eval_mar_100": 0.5823,
601
+ "eval_mar_100_bus": 0.6637,
602
+ "eval_mar_100_car": 0.591,
603
+ "eval_mar_100_container": 0.6308,
604
+ "eval_mar_100_motorbike": 0.4437,
605
+ "eval_mar_large": 0.7735,
606
+ "eval_mar_medium": 0.5844,
607
+ "eval_mar_small": 0.2885,
608
+ "eval_runtime": 144.7391,
609
+ "eval_samples_per_second": 15.911,
610
+ "eval_steps_per_second": 1.99,
611
+ "step": 23050
612
+ },
613
+ {
614
+ "epoch": 10.195227765726681,
615
+ "grad_norm": 6654.49365234375,
616
+ "learning_rate": 9.30532812271677e-05,
617
+ "loss": 0.9821,
618
+ "step": 23500
619
+ },
620
+ {
621
+ "epoch": 10.412147505422993,
622
+ "grad_norm": 12540.853515625,
623
+ "learning_rate": 9.276173960919273e-05,
624
+ "loss": 0.9966,
625
+ "step": 24000
626
+ },
627
+ {
628
+ "epoch": 10.629067245119305,
629
+ "grad_norm": 8583.029296875,
630
+ "learning_rate": 9.246468171391835e-05,
631
+ "loss": 0.9826,
632
+ "step": 24500
633
+ },
634
+ {
635
+ "epoch": 10.845986984815617,
636
+ "grad_norm": 5945.16796875,
637
+ "learning_rate": 9.216214586190261e-05,
638
+ "loss": 0.9769,
639
+ "step": 25000
640
+ },
641
+ {
642
+ "epoch": 11.0,
643
+ "eval_loss": 0.9336058497428894,
644
+ "eval_map": 0.4511,
645
+ "eval_map_50": 0.8,
646
+ "eval_map_75": 0.4575,
647
+ "eval_map_bus": 0.5073,
648
+ "eval_map_car": 0.4708,
649
+ "eval_map_container": 0.5006,
650
+ "eval_map_large": 0.6611,
651
+ "eval_map_medium": 0.4469,
652
+ "eval_map_motorbike": 0.3259,
653
+ "eval_map_small": 0.1072,
654
+ "eval_mar_1": 0.2412,
655
+ "eval_mar_10": 0.5266,
656
+ "eval_mar_100": 0.5752,
657
+ "eval_mar_100_bus": 0.6493,
658
+ "eval_mar_100_car": 0.5772,
659
+ "eval_mar_100_container": 0.6233,
660
+ "eval_mar_100_motorbike": 0.4511,
661
+ "eval_mar_large": 0.7508,
662
+ "eval_mar_medium": 0.5789,
663
+ "eval_mar_small": 0.2908,
664
+ "eval_runtime": 147.6126,
665
+ "eval_samples_per_second": 15.602,
666
+ "eval_steps_per_second": 1.951,
667
+ "step": 25355
668
+ },
669
+ {
670
+ "epoch": 11.062906724511931,
671
+ "grad_norm": 10052.8896484375,
672
+ "learning_rate": 9.185417108036163e-05,
673
+ "loss": 0.9731,
674
+ "step": 25500
675
+ },
676
+ {
677
+ "epoch": 11.279826464208243,
678
+ "grad_norm": 3725.647216796875,
679
+ "learning_rate": 9.154079709813505e-05,
680
+ "loss": 0.9572,
681
+ "step": 26000
682
+ },
683
+ {
684
+ "epoch": 11.496746203904555,
685
+ "grad_norm": 10829.8896484375,
686
+ "learning_rate": 9.122206434056104e-05,
687
+ "loss": 0.9558,
688
+ "step": 26500
689
+ },
690
+ {
691
+ "epoch": 11.713665943600867,
692
+ "grad_norm": 16212.732421875,
693
+ "learning_rate": 9.08980139242614e-05,
694
+ "loss": 0.9622,
695
+ "step": 27000
696
+ },
697
+ {
698
+ "epoch": 11.93058568329718,
699
+ "grad_norm": 12326.623046875,
700
+ "learning_rate": 9.056868765183752e-05,
701
+ "loss": 0.9506,
702
+ "step": 27500
703
+ },
704
+ {
705
+ "epoch": 12.0,
706
+ "eval_loss": 0.9213413596153259,
707
+ "eval_map": 0.4617,
708
+ "eval_map_50": 0.8095,
709
+ "eval_map_75": 0.4735,
710
+ "eval_map_bus": 0.5252,
711
+ "eval_map_car": 0.4891,
712
+ "eval_map_container": 0.5052,
713
+ "eval_map_large": 0.676,
714
+ "eval_map_medium": 0.455,
715
+ "eval_map_motorbike": 0.3272,
716
+ "eval_map_small": 0.1083,
717
+ "eval_mar_1": 0.2461,
718
+ "eval_mar_10": 0.5327,
719
+ "eval_mar_100": 0.5818,
720
+ "eval_mar_100_bus": 0.6626,
721
+ "eval_mar_100_car": 0.5902,
722
+ "eval_mar_100_container": 0.627,
723
+ "eval_mar_100_motorbike": 0.4475,
724
+ "eval_mar_large": 0.7631,
725
+ "eval_mar_medium": 0.5838,
726
+ "eval_mar_small": 0.2935,
727
+ "eval_runtime": 143.9851,
728
+ "eval_samples_per_second": 15.995,
729
+ "eval_steps_per_second": 2.0,
730
+ "step": 27660
731
+ },
732
+ {
733
+ "epoch": 12.147505422993492,
734
+ "grad_norm": 23455.04296875,
735
+ "learning_rate": 9.023412800647782e-05,
736
+ "loss": 0.9338,
737
+ "step": 28000
738
+ },
739
+ {
740
+ "epoch": 12.364425162689805,
741
+ "grad_norm": 7761.9912109375,
742
+ "learning_rate": 8.989506279701155e-05,
743
+ "loss": 0.9535,
744
+ "step": 28500
745
+ },
746
+ {
747
+ "epoch": 12.581344902386117,
748
+ "grad_norm": 7034.04052734375,
749
+ "learning_rate": 8.955017679879593e-05,
750
+ "loss": 0.9507,
751
+ "step": 29000
752
+ },
753
+ {
754
+ "epoch": 12.79826464208243,
755
+ "grad_norm": 5548.38525390625,
756
+ "learning_rate": 8.920018881585156e-05,
757
+ "loss": 0.9417,
758
+ "step": 29500
759
+ },
760
+ {
761
+ "epoch": 13.0,
762
+ "eval_loss": 0.9276334643363953,
763
+ "eval_map": 0.4608,
764
+ "eval_map_50": 0.8076,
765
+ "eval_map_75": 0.4687,
766
+ "eval_map_bus": 0.5336,
767
+ "eval_map_car": 0.4866,
768
+ "eval_map_container": 0.4992,
769
+ "eval_map_large": 0.6724,
770
+ "eval_map_medium": 0.4511,
771
+ "eval_map_motorbike": 0.3238,
772
+ "eval_map_small": 0.1064,
773
+ "eval_mar_1": 0.2472,
774
+ "eval_mar_10": 0.5288,
775
+ "eval_mar_100": 0.5752,
776
+ "eval_mar_100_bus": 0.6588,
777
+ "eval_mar_100_car": 0.5895,
778
+ "eval_mar_100_container": 0.6158,
779
+ "eval_mar_100_motorbike": 0.4369,
780
+ "eval_mar_large": 0.7512,
781
+ "eval_mar_medium": 0.5708,
782
+ "eval_mar_small": 0.2981,
783
+ "eval_runtime": 145.1236,
784
+ "eval_samples_per_second": 15.869,
785
+ "eval_steps_per_second": 1.985,
786
+ "step": 29965
787
+ },
788
+ {
789
+ "epoch": 13.015184381778742,
790
+ "grad_norm": 12253.849609375,
791
+ "learning_rate": 8.884514399673369e-05,
792
+ "loss": 0.9555,
793
+ "step": 30000
794
+ },
795
+ {
796
+ "epoch": 13.232104121475054,
797
+ "grad_norm": 13796.53125,
798
+ "learning_rate": 8.848508814233106e-05,
799
+ "loss": 0.9476,
800
+ "step": 30500
801
+ },
802
+ {
803
+ "epoch": 13.449023861171366,
804
+ "grad_norm": 5713.0361328125,
805
+ "learning_rate": 8.812080266430921e-05,
806
+ "loss": 0.9417,
807
+ "step": 31000
808
+ },
809
+ {
810
+ "epoch": 13.66594360086768,
811
+ "grad_norm": 3508.00634765625,
812
+ "learning_rate": 8.77508745092958e-05,
813
+ "loss": 0.9446,
814
+ "step": 31500
815
+ },
816
+ {
817
+ "epoch": 13.882863340563992,
818
+ "grad_norm": 4413.595703125,
819
+ "learning_rate": 8.737607648009131e-05,
820
+ "loss": 0.9488,
821
+ "step": 32000
822
+ },
823
+ {
824
+ "epoch": 14.0,
825
+ "eval_loss": 0.9241105318069458,
826
+ "eval_map": 0.4602,
827
+ "eval_map_50": 0.8064,
828
+ "eval_map_75": 0.4696,
829
+ "eval_map_bus": 0.5309,
830
+ "eval_map_car": 0.4761,
831
+ "eval_map_container": 0.5062,
832
+ "eval_map_large": 0.6817,
833
+ "eval_map_medium": 0.4503,
834
+ "eval_map_motorbike": 0.3275,
835
+ "eval_map_small": 0.1122,
836
+ "eval_mar_1": 0.2462,
837
+ "eval_mar_10": 0.5333,
838
+ "eval_mar_100": 0.5819,
839
+ "eval_mar_100_bus": 0.6633,
840
+ "eval_mar_100_car": 0.5828,
841
+ "eval_mar_100_container": 0.6259,
842
+ "eval_mar_100_motorbike": 0.4557,
843
+ "eval_mar_large": 0.7655,
844
+ "eval_mar_medium": 0.5804,
845
+ "eval_mar_small": 0.2896,
846
+ "eval_runtime": 145.7229,
847
+ "eval_samples_per_second": 15.804,
848
+ "eval_steps_per_second": 1.976,
849
+ "step": 32270
850
+ },
851
+ {
852
+ "epoch": 14.099783080260304,
853
+ "grad_norm": 7677.90478515625,
854
+ "learning_rate": 8.699645692575454e-05,
855
+ "loss": 0.9519,
856
+ "step": 32500
857
+ },
858
+ {
859
+ "epoch": 14.316702819956616,
860
+ "grad_norm": 4366.18212890625,
861
+ "learning_rate": 8.661283833169138e-05,
862
+ "loss": 0.9471,
863
+ "step": 33000
864
+ },
865
+ {
866
+ "epoch": 14.533622559652928,
867
+ "grad_norm": 5411.373046875,
868
+ "learning_rate": 8.622373265190294e-05,
869
+ "loss": 0.9423,
870
+ "step": 33500
871
+ },
872
+ {
873
+ "epoch": 14.750542299349242,
874
+ "grad_norm": 7791.236328125,
875
+ "learning_rate": 8.582995409968335e-05,
876
+ "loss": 0.9357,
877
+ "step": 34000
878
+ },
879
+ {
880
+ "epoch": 14.967462039045554,
881
+ "grad_norm": 20878.41796875,
882
+ "learning_rate": 8.543155347258457e-05,
883
+ "loss": 0.9371,
884
+ "step": 34500
885
+ },
886
+ {
887
+ "epoch": 15.0,
888
+ "eval_loss": 0.9158653020858765,
889
+ "eval_map": 0.4646,
890
+ "eval_map_50": 0.8109,
891
+ "eval_map_75": 0.479,
892
+ "eval_map_bus": 0.525,
893
+ "eval_map_car": 0.4942,
894
+ "eval_map_container": 0.5118,
895
+ "eval_map_large": 0.6799,
896
+ "eval_map_medium": 0.4567,
897
+ "eval_map_motorbike": 0.3276,
898
+ "eval_map_small": 0.1141,
899
+ "eval_mar_1": 0.2484,
900
+ "eval_mar_10": 0.533,
901
+ "eval_mar_100": 0.5795,
902
+ "eval_mar_100_bus": 0.6523,
903
+ "eval_mar_100_car": 0.5909,
904
+ "eval_mar_100_container": 0.6278,
905
+ "eval_mar_100_motorbike": 0.447,
906
+ "eval_mar_large": 0.7745,
907
+ "eval_mar_medium": 0.5767,
908
+ "eval_mar_small": 0.3179,
909
+ "eval_runtime": 143.8949,
910
+ "eval_samples_per_second": 16.005,
911
+ "eval_steps_per_second": 2.001,
912
+ "step": 34575
913
+ },
914
+ {
915
+ "epoch": 15.184381778741866,
916
+ "grad_norm": 8804.869140625,
917
+ "learning_rate": 8.502858216440758e-05,
918
+ "loss": 0.9249,
919
+ "step": 35000
920
+ },
921
+ {
922
+ "epoch": 15.401301518438178,
923
+ "grad_norm": 7941.79443359375,
924
+ "learning_rate": 8.462191161340177e-05,
925
+ "loss": 0.9341,
926
+ "step": 35500
927
+ },
928
+ {
929
+ "epoch": 15.61822125813449,
930
+ "grad_norm": 11279.1513671875,
931
+ "learning_rate": 8.420996435565979e-05,
932
+ "loss": 0.9342,
933
+ "step": 36000
934
+ },
935
+ {
936
+ "epoch": 15.835140997830802,
937
+ "grad_norm": 2519.469970703125,
938
+ "learning_rate": 8.379360400220391e-05,
939
+ "loss": 0.9311,
940
+ "step": 36500
941
+ },
942
+ {
943
+ "epoch": 16.0,
944
+ "eval_loss": 0.8999801278114319,
945
+ "eval_map": 0.4741,
946
+ "eval_map_50": 0.8169,
947
+ "eval_map_75": 0.4893,
948
+ "eval_map_bus": 0.5392,
949
+ "eval_map_car": 0.5006,
950
+ "eval_map_container": 0.5198,
951
+ "eval_map_large": 0.6914,
952
+ "eval_map_medium": 0.4673,
953
+ "eval_map_motorbike": 0.3368,
954
+ "eval_map_small": 0.1224,
955
+ "eval_mar_1": 0.2523,
956
+ "eval_mar_10": 0.544,
957
+ "eval_mar_100": 0.5909,
958
+ "eval_mar_100_bus": 0.6792,
959
+ "eval_mar_100_car": 0.5917,
960
+ "eval_mar_100_container": 0.6356,
961
+ "eval_mar_100_motorbike": 0.4571,
962
+ "eval_mar_large": 0.7817,
963
+ "eval_mar_medium": 0.593,
964
+ "eval_mar_small": 0.3192,
965
+ "eval_runtime": 144.4917,
966
+ "eval_samples_per_second": 15.939,
967
+ "eval_steps_per_second": 1.993,
968
+ "step": 36880
969
+ },
970
+ {
971
+ "epoch": 16.052060737527114,
972
+ "grad_norm": 6870.65673828125,
973
+ "learning_rate": 8.337288426364522e-05,
974
+ "loss": 0.9394,
975
+ "step": 37000
976
+ },
977
+ {
978
+ "epoch": 16.268980477223426,
979
+ "grad_norm": 18748.70703125,
980
+ "learning_rate": 8.294785941295685e-05,
981
+ "loss": 0.9206,
982
+ "step": 37500
983
+ },
984
+ {
985
+ "epoch": 16.485900216919738,
986
+ "grad_norm": 8770.451171875,
987
+ "learning_rate": 8.251944703380834e-05,
988
+ "loss": 0.9277,
989
+ "step": 38000
990
+ },
991
+ {
992
+ "epoch": 16.702819956616054,
993
+ "grad_norm": 10527.71484375,
994
+ "learning_rate": 8.208598532633761e-05,
995
+ "loss": 0.9363,
996
+ "step": 38500
997
+ },
998
+ {
999
+ "epoch": 16.919739696312366,
1000
+ "grad_norm": 13457.677734375,
1001
+ "learning_rate": 8.164838451708801e-05,
1002
+ "loss": 0.9284,
1003
+ "step": 39000
1004
+ },
1005
+ {
1006
+ "epoch": 17.0,
1007
+ "eval_loss": 0.9076481461524963,
1008
+ "eval_map": 0.472,
1009
+ "eval_map_50": 0.8148,
1010
+ "eval_map_75": 0.4845,
1011
+ "eval_map_bus": 0.5397,
1012
+ "eval_map_car": 0.4969,
1013
+ "eval_map_container": 0.5162,
1014
+ "eval_map_large": 0.684,
1015
+ "eval_map_medium": 0.4625,
1016
+ "eval_map_motorbike": 0.3354,
1017
+ "eval_map_small": 0.1174,
1018
+ "eval_mar_1": 0.252,
1019
+ "eval_mar_10": 0.5415,
1020
+ "eval_mar_100": 0.5905,
1021
+ "eval_mar_100_bus": 0.6732,
1022
+ "eval_mar_100_car": 0.5938,
1023
+ "eval_mar_100_container": 0.6354,
1024
+ "eval_mar_100_motorbike": 0.4596,
1025
+ "eval_mar_large": 0.767,
1026
+ "eval_mar_medium": 0.5905,
1027
+ "eval_mar_small": 0.315,
1028
+ "eval_runtime": 145.0152,
1029
+ "eval_samples_per_second": 15.881,
1030
+ "eval_steps_per_second": 1.986,
1031
+ "step": 39185
1032
+ },
1033
+ {
1034
+ "epoch": 17.136659436008678,
1035
+ "grad_norm": 6826.60595703125,
1036
+ "learning_rate": 8.120670105669583e-05,
1037
+ "loss": 0.9237,
1038
+ "step": 39500
1039
+ },
1040
+ {
1041
+ "epoch": 17.35357917570499,
1042
+ "grad_norm": 10753.5400390625,
1043
+ "learning_rate": 8.076099192246061e-05,
1044
+ "loss": 0.9257,
1045
+ "step": 40000
1046
+ },
1047
+ {
1048
+ "epoch": 17.570498915401302,
1049
+ "grad_norm": 5758.92919921875,
1050
+ "learning_rate": 8.031131461099501e-05,
1051
+ "loss": 0.9247,
1052
+ "step": 40500
1053
+ },
1054
+ {
1055
+ "epoch": 17.787418655097614,
1056
+ "grad_norm": 11779.173828125,
1057
+ "learning_rate": 7.985772713080777e-05,
1058
+ "loss": 0.9107,
1059
+ "step": 41000
1060
+ },
1061
+ {
1062
+ "epoch": 18.0,
1063
+ "eval_loss": 0.9001684188842773,
1064
+ "eval_map": 0.4753,
1065
+ "eval_map_50": 0.8198,
1066
+ "eval_map_75": 0.4952,
1067
+ "eval_map_bus": 0.5451,
1068
+ "eval_map_car": 0.501,
1069
+ "eval_map_container": 0.5198,
1070
+ "eval_map_large": 0.6944,
1071
+ "eval_map_medium": 0.4679,
1072
+ "eval_map_motorbike": 0.3352,
1073
+ "eval_map_small": 0.1207,
1074
+ "eval_mar_1": 0.2516,
1075
+ "eval_mar_10": 0.5428,
1076
+ "eval_mar_100": 0.5902,
1077
+ "eval_mar_100_bus": 0.6716,
1078
+ "eval_mar_100_car": 0.5977,
1079
+ "eval_mar_100_container": 0.6353,
1080
+ "eval_mar_100_motorbike": 0.4562,
1081
+ "eval_mar_large": 0.7743,
1082
+ "eval_mar_medium": 0.592,
1083
+ "eval_mar_small": 0.3133,
1084
+ "eval_runtime": 143.7248,
1085
+ "eval_samples_per_second": 16.024,
1086
+ "eval_steps_per_second": 2.004,
1087
+ "step": 41490
1088
+ },
1089
+ {
1090
+ "epoch": 18.004338394793926,
1091
+ "grad_norm": 3542.92626953125,
1092
+ "learning_rate": 7.940120667790506e-05,
1093
+ "loss": 0.919,
1094
+ "step": 41500
1095
+ },
1096
+ {
1097
+ "epoch": 18.22125813449024,
1098
+ "grad_norm": 22165.880859375,
1099
+ "learning_rate": 7.893998242197806e-05,
1100
+ "loss": 0.928,
1101
+ "step": 42000
1102
+ },
1103
+ {
1104
+ "epoch": 18.43817787418655,
1105
+ "grad_norm": 3429.292236328125,
1106
+ "learning_rate": 7.84750248995952e-05,
1107
+ "loss": 0.9192,
1108
+ "step": 42500
1109
+ },
1110
+ {
1111
+ "epoch": 18.655097613882862,
1112
+ "grad_norm": 10154.87109375,
1113
+ "learning_rate": 7.800639409041694e-05,
1114
+ "loss": 0.9183,
1115
+ "step": 43000
1116
+ },
1117
+ {
1118
+ "epoch": 18.872017353579174,
1119
+ "grad_norm": 25265.55078125,
1120
+ "learning_rate": 7.753415044795898e-05,
1121
+ "loss": 0.9125,
1122
+ "step": 43500
1123
+ },
1124
+ {
1125
+ "epoch": 19.0,
1126
+ "eval_loss": 0.8902826905250549,
1127
+ "eval_map": 0.4805,
1128
+ "eval_map_50": 0.8223,
1129
+ "eval_map_75": 0.5006,
1130
+ "eval_map_bus": 0.5537,
1131
+ "eval_map_car": 0.5047,
1132
+ "eval_map_container": 0.5198,
1133
+ "eval_map_large": 0.6876,
1134
+ "eval_map_medium": 0.4736,
1135
+ "eval_map_motorbike": 0.3438,
1136
+ "eval_map_small": 0.119,
1137
+ "eval_mar_1": 0.2528,
1138
+ "eval_mar_10": 0.5449,
1139
+ "eval_mar_100": 0.5921,
1140
+ "eval_mar_100_bus": 0.6688,
1141
+ "eval_mar_100_car": 0.6008,
1142
+ "eval_mar_100_container": 0.6319,
1143
+ "eval_mar_100_motorbike": 0.4669,
1144
+ "eval_mar_large": 0.7716,
1145
+ "eval_mar_medium": 0.5905,
1146
+ "eval_mar_small": 0.3164,
1147
+ "eval_runtime": 146.715,
1148
+ "eval_samples_per_second": 15.697,
1149
+ "eval_steps_per_second": 1.963,
1150
+ "step": 43795
1151
+ },
1152
+ {
1153
+ "epoch": 19.088937093275486,
1154
+ "grad_norm": 7347.31640625,
1155
+ "learning_rate": 7.705835489179352e-05,
1156
+ "loss": 0.9033,
1157
+ "step": 44000
1158
+ },
1159
+ {
1160
+ "epoch": 19.305856832971802,
1161
+ "grad_norm": 5815.42822265625,
1162
+ "learning_rate": 7.658003081440886e-05,
1163
+ "loss": 0.9049,
1164
+ "step": 44500
1165
+ },
1166
+ {
1167
+ "epoch": 19.522776572668114,
1168
+ "grad_norm": 5560.4921875,
1169
+ "learning_rate": 7.609732280983515e-05,
1170
+ "loss": 0.9124,
1171
+ "step": 45000
1172
+ },
1173
+ {
1174
+ "epoch": 19.739696312364426,
1175
+ "grad_norm": 5130.99462890625,
1176
+ "learning_rate": 7.56112482427524e-05,
1177
+ "loss": 0.9205,
1178
+ "step": 45500
1179
+ },
1180
+ {
1181
+ "epoch": 19.95661605206074,
1182
+ "grad_norm": 17505.642578125,
1183
+ "learning_rate": 7.512186981692624e-05,
1184
+ "loss": 0.913,
1185
+ "step": 46000
1186
+ },
1187
+ {
1188
+ "epoch": 20.0,
1189
+ "eval_loss": 0.8936046957969666,
1190
+ "eval_map": 0.4822,
1191
+ "eval_map_50": 0.8243,
1192
+ "eval_map_75": 0.504,
1193
+ "eval_map_bus": 0.5592,
1194
+ "eval_map_car": 0.4984,
1195
+ "eval_map_container": 0.5236,
1196
+ "eval_map_large": 0.6904,
1197
+ "eval_map_medium": 0.4777,
1198
+ "eval_map_motorbike": 0.3475,
1199
+ "eval_map_small": 0.115,
1200
+ "eval_mar_1": 0.2536,
1201
+ "eval_mar_10": 0.5453,
1202
+ "eval_mar_100": 0.5923,
1203
+ "eval_mar_100_bus": 0.6727,
1204
+ "eval_mar_100_car": 0.5953,
1205
+ "eval_mar_100_container": 0.6347,
1206
+ "eval_mar_100_motorbike": 0.4666,
1207
+ "eval_mar_large": 0.7683,
1208
+ "eval_mar_medium": 0.5946,
1209
+ "eval_mar_small": 0.3124,
1210
+ "eval_runtime": 146.1172,
1211
+ "eval_samples_per_second": 15.761,
1212
+ "eval_steps_per_second": 1.971,
1213
+ "step": 46100
1214
+ },
1215
+ {
1216
+ "epoch": 20.17353579175705,
1217
+ "grad_norm": 21176.666015625,
1218
+ "learning_rate": 7.462925066232113e-05,
1219
+ "loss": 0.9039,
1220
+ "step": 46500
1221
+ },
1222
+ {
1223
+ "epoch": 20.390455531453362,
1224
+ "grad_norm": 4085.164306640625,
1225
+ "learning_rate": 7.413444904801033e-05,
1226
+ "loss": 0.9082,
1227
+ "step": 47000
1228
+ },
1229
+ {
1230
+ "epoch": 20.607375271149674,
1231
+ "grad_norm": 20771.896484375,
1232
+ "learning_rate": 7.363554565211042e-05,
1233
+ "loss": 0.9087,
1234
+ "step": 47500
1235
+ },
1236
+ {
1237
+ "epoch": 20.824295010845987,
1238
+ "grad_norm": 5990.69580078125,
1239
+ "learning_rate": 7.313359326369684e-05,
1240
+ "loss": 0.908,
1241
+ "step": 48000
1242
+ },
1243
+ {
1244
+ "epoch": 21.0,
1245
+ "eval_loss": 0.8810420036315918,
1246
+ "eval_map": 0.483,
1247
+ "eval_map_50": 0.8249,
1248
+ "eval_map_75": 0.5047,
1249
+ "eval_map_bus": 0.5531,
1250
+ "eval_map_car": 0.5081,
1251
+ "eval_map_container": 0.5229,
1252
+ "eval_map_large": 0.7091,
1253
+ "eval_map_medium": 0.4743,
1254
+ "eval_map_motorbike": 0.348,
1255
+ "eval_map_small": 0.1233,
1256
+ "eval_mar_1": 0.2539,
1257
+ "eval_mar_10": 0.5486,
1258
+ "eval_mar_100": 0.5987,
1259
+ "eval_mar_100_bus": 0.6764,
1260
+ "eval_mar_100_car": 0.6079,
1261
+ "eval_mar_100_container": 0.6426,
1262
+ "eval_mar_100_motorbike": 0.4679,
1263
+ "eval_mar_large": 0.7947,
1264
+ "eval_mar_medium": 0.5992,
1265
+ "eval_mar_small": 0.3117,
1266
+ "eval_runtime": 143.798,
1267
+ "eval_samples_per_second": 16.016,
1268
+ "eval_steps_per_second": 2.003,
1269
+ "step": 48405
1270
+ },
1271
+ {
1272
+ "epoch": 21.0412147505423,
1273
+ "grad_norm": 4679.26220703125,
1274
+ "learning_rate": 7.262865663477902e-05,
1275
+ "loss": 0.9023,
1276
+ "step": 48500
1277
+ },
1278
+ {
1279
+ "epoch": 21.25813449023861,
1280
+ "grad_norm": 18654.177734375,
1281
+ "learning_rate": 7.21208009023343e-05,
1282
+ "loss": 0.9069,
1283
+ "step": 49000
1284
+ },
1285
+ {
1286
+ "epoch": 21.475054229934923,
1287
+ "grad_norm": 10939.3623046875,
1288
+ "learning_rate": 7.161009157990521e-05,
1289
+ "loss": 0.8956,
1290
+ "step": 49500
1291
+ },
1292
+ {
1293
+ "epoch": 21.691973969631235,
1294
+ "grad_norm": 14381.9765625,
1295
+ "learning_rate": 7.109659454914833e-05,
1296
+ "loss": 0.8925,
1297
+ "step": 50000
1298
+ },
1299
+ {
1300
+ "epoch": 21.90889370932755,
1301
+ "grad_norm": 5005.86865234375,
1302
+ "learning_rate": 7.058141116015095e-05,
1303
+ "loss": 0.8997,
1304
+ "step": 50500
1305
+ },
1306
+ {
1307
+ "epoch": 22.0,
1308
+ "eval_loss": 0.8783358335494995,
1309
+ "eval_map": 0.4846,
1310
+ "eval_map_50": 0.8275,
1311
+ "eval_map_75": 0.5024,
1312
+ "eval_map_bus": 0.5507,
1313
+ "eval_map_car": 0.5086,
1314
+ "eval_map_container": 0.5299,
1315
+ "eval_map_large": 0.699,
1316
+ "eval_map_medium": 0.4761,
1317
+ "eval_map_motorbike": 0.3491,
1318
+ "eval_map_small": 0.1256,
1319
+ "eval_mar_1": 0.2542,
1320
+ "eval_mar_10": 0.5503,
1321
+ "eval_mar_100": 0.5994,
1322
+ "eval_mar_100_bus": 0.6731,
1323
+ "eval_mar_100_car": 0.6068,
1324
+ "eval_mar_100_container": 0.6486,
1325
+ "eval_mar_100_motorbike": 0.469,
1326
+ "eval_mar_large": 0.7854,
1327
+ "eval_mar_medium": 0.5993,
1328
+ "eval_mar_small": 0.3314,
1329
+ "eval_runtime": 144.6256,
1330
+ "eval_samples_per_second": 15.924,
1331
+ "eval_steps_per_second": 1.991,
1332
+ "step": 50710
1333
+ },
1334
+ {
1335
+ "epoch": 22.125813449023862,
1336
+ "grad_norm": 19755.140625,
1337
+ "learning_rate": 7.006254303068497e-05,
1338
+ "loss": 0.8975,
1339
+ "step": 51000
1340
+ },
1341
+ {
1342
+ "epoch": 22.342733188720175,
1343
+ "grad_norm": 11582.6494140625,
1344
+ "learning_rate": 6.954108682712056e-05,
1345
+ "loss": 0.898,
1346
+ "step": 51500
1347
+ },
1348
+ {
1349
+ "epoch": 22.559652928416487,
1350
+ "grad_norm": 6819.56396484375,
1351
+ "learning_rate": 6.901710981746509e-05,
1352
+ "loss": 0.8999,
1353
+ "step": 52000
1354
+ },
1355
+ {
1356
+ "epoch": 22.7765726681128,
1357
+ "grad_norm": 7609.8115234375,
1358
+ "learning_rate": 6.849067959491079e-05,
1359
+ "loss": 0.8978,
1360
+ "step": 52500
1361
+ },
1362
+ {
1363
+ "epoch": 22.99349240780911,
1364
+ "grad_norm": 10856.244140625,
1365
+ "learning_rate": 6.796186406911505e-05,
1366
+ "loss": 0.8865,
1367
+ "step": 53000
1368
+ },
1369
+ {
1370
+ "epoch": 23.0,
1371
+ "eval_loss": 0.871816098690033,
1372
+ "eval_map": 0.4901,
1373
+ "eval_map_50": 0.8289,
1374
+ "eval_map_75": 0.5202,
1375
+ "eval_map_bus": 0.5591,
1376
+ "eval_map_car": 0.5153,
1377
+ "eval_map_container": 0.5306,
1378
+ "eval_map_large": 0.7037,
1379
+ "eval_map_medium": 0.4841,
1380
+ "eval_map_motorbike": 0.3553,
1381
+ "eval_map_small": 0.1243,
1382
+ "eval_mar_1": 0.2558,
1383
+ "eval_mar_10": 0.5526,
1384
+ "eval_mar_100": 0.6008,
1385
+ "eval_mar_100_bus": 0.6757,
1386
+ "eval_mar_100_car": 0.6077,
1387
+ "eval_mar_100_container": 0.6452,
1388
+ "eval_mar_100_motorbike": 0.4746,
1389
+ "eval_mar_large": 0.788,
1390
+ "eval_mar_medium": 0.6019,
1391
+ "eval_mar_small": 0.3201,
1392
+ "eval_runtime": 145.8422,
1393
+ "eval_samples_per_second": 15.791,
1394
+ "eval_steps_per_second": 1.975,
1395
+ "step": 53015
1396
+ },
1397
+ {
1398
+ "epoch": 23.210412147505423,
1399
+ "grad_norm": 6373.787109375,
1400
+ "learning_rate": 6.743286051281191e-05,
1401
+ "loss": 0.8974,
1402
+ "step": 53500
1403
+ },
1404
+ {
1405
+ "epoch": 23.427331887201735,
1406
+ "grad_norm": 3711.38671875,
1407
+ "learning_rate": 6.68994881889314e-05,
1408
+ "loss": 0.9048,
1409
+ "step": 54000
1410
+ },
1411
+ {
1412
+ "epoch": 23.644251626898047,
1413
+ "grad_norm": 1683.364501953125,
1414
+ "learning_rate": 6.636393582598046e-05,
1415
+ "loss": 0.911,
1416
+ "step": 54500
1417
+ },
1418
+ {
1419
+ "epoch": 23.86117136659436,
1420
+ "grad_norm": 2381.84423828125,
1421
+ "learning_rate": 6.582627251037533e-05,
1422
+ "loss": 0.9052,
1423
+ "step": 55000
1424
+ },
1425
+ {
1426
+ "epoch": 24.0,
1427
+ "eval_loss": 0.8951303958892822,
1428
+ "eval_map": 0.4779,
1429
+ "eval_map_50": 0.8254,
1430
+ "eval_map_75": 0.4914,
1431
+ "eval_map_bus": 0.5509,
1432
+ "eval_map_car": 0.5002,
1433
+ "eval_map_container": 0.5227,
1434
+ "eval_map_large": 0.697,
1435
+ "eval_map_medium": 0.469,
1436
+ "eval_map_motorbike": 0.3376,
1437
+ "eval_map_small": 0.1198,
1438
+ "eval_mar_1": 0.2524,
1439
+ "eval_mar_10": 0.5426,
1440
+ "eval_mar_100": 0.5902,
1441
+ "eval_mar_100_bus": 0.6715,
1442
+ "eval_mar_100_car": 0.5914,
1443
+ "eval_mar_100_container": 0.643,
1444
+ "eval_mar_100_motorbike": 0.455,
1445
+ "eval_mar_large": 0.7785,
1446
+ "eval_mar_medium": 0.5885,
1447
+ "eval_mar_small": 0.319,
1448
+ "eval_runtime": 145.8551,
1449
+ "eval_samples_per_second": 15.79,
1450
+ "eval_steps_per_second": 1.975,
1451
+ "step": 55320
1452
+ },
1453
+ {
1454
+ "epoch": 24.07809110629067,
1455
+ "grad_norm": 19113.990234375,
1456
+ "learning_rate": 6.528656760084578e-05,
1457
+ "loss": 0.9112,
1458
+ "step": 55500
1459
+ },
1460
+ {
1461
+ "epoch": 24.295010845986983,
1462
+ "grad_norm": 20675.185546875,
1463
+ "learning_rate": 6.474489071948787e-05,
1464
+ "loss": 0.8944,
1465
+ "step": 56000
1466
+ },
1467
+ {
1468
+ "epoch": 24.5119305856833,
1469
+ "grad_norm": 10454.0859375,
1470
+ "learning_rate": 6.420131174278254e-05,
1471
+ "loss": 0.8915,
1472
+ "step": 56500
1473
+ },
1474
+ {
1475
+ "epoch": 24.72885032537961,
1476
+ "grad_norm": 4770.7431640625,
1477
+ "learning_rate": 6.365590079258165e-05,
1478
+ "loss": 0.8891,
1479
+ "step": 57000
1480
+ },
1481
+ {
1482
+ "epoch": 24.945770065075923,
1483
+ "grad_norm": 16598.36328125,
1484
+ "learning_rate": 6.310872822706217e-05,
1485
+ "loss": 0.8982,
1486
+ "step": 57500
1487
+ },
1488
+ {
1489
+ "epoch": 25.0,
1490
+ "eval_loss": 0.8864476084709167,
1491
+ "eval_map": 0.4662,
1492
+ "eval_map_50": 0.8105,
1493
+ "eval_map_75": 0.4884,
1494
+ "eval_map_bus": 0.515,
1495
+ "eval_map_car": 0.4889,
1496
+ "eval_map_container": 0.5084,
1497
+ "eval_map_large": 0.6737,
1498
+ "eval_map_medium": 0.4711,
1499
+ "eval_map_motorbike": 0.3524,
1500
+ "eval_map_small": 0.1286,
1501
+ "eval_mar_1": 0.2482,
1502
+ "eval_mar_10": 0.5444,
1503
+ "eval_mar_100": 0.5934,
1504
+ "eval_mar_100_bus": 0.6696,
1505
+ "eval_mar_100_car": 0.5993,
1506
+ "eval_mar_100_container": 0.6371,
1507
+ "eval_mar_100_motorbike": 0.4675,
1508
+ "eval_mar_large": 0.7729,
1509
+ "eval_mar_medium": 0.5969,
1510
+ "eval_mar_small": 0.3226,
1511
+ "eval_runtime": 144.7854,
1512
+ "eval_samples_per_second": 15.906,
1513
+ "eval_steps_per_second": 1.989,
1514
+ "step": 57625
1515
+ },
1516
+ {
1517
+ "epoch": 25.162689804772235,
1518
+ "grad_norm": 6422.17041015625,
1519
+ "learning_rate": 6.25609639994632e-05,
1520
+ "loss": 0.8784,
1521
+ "step": 58000
1522
+ },
1523
+ {
1524
+ "epoch": 25.379609544468547,
1525
+ "grad_norm": 16689.212890625,
1526
+ "learning_rate": 6.201048334737784e-05,
1527
+ "loss": 0.8855,
1528
+ "step": 58500
1529
+ },
1530
+ {
1531
+ "epoch": 25.59652928416486,
1532
+ "grad_norm": 18894.322265625,
1533
+ "learning_rate": 6.145845333932029e-05,
1534
+ "loss": 0.8872,
1535
+ "step": 59000
1536
+ },
1537
+ {
1538
+ "epoch": 25.81344902386117,
1539
+ "grad_norm": 4359.09521484375,
1540
+ "learning_rate": 6.090494518732801e-05,
1541
+ "loss": 0.8835,
1542
+ "step": 59500
1543
+ },
1544
+ {
1545
+ "epoch": 26.0,
1546
+ "eval_loss": 0.8596058487892151,
1547
+ "eval_map": 0.4972,
1548
+ "eval_map_50": 0.8376,
1549
+ "eval_map_75": 0.5257,
1550
+ "eval_map_bus": 0.5714,
1551
+ "eval_map_car": 0.5186,
1552
+ "eval_map_container": 0.5391,
1553
+ "eval_map_large": 0.7028,
1554
+ "eval_map_medium": 0.4915,
1555
+ "eval_map_motorbike": 0.3598,
1556
+ "eval_map_small": 0.1342,
1557
+ "eval_mar_1": 0.2583,
1558
+ "eval_mar_10": 0.5574,
1559
+ "eval_mar_100": 0.6057,
1560
+ "eval_mar_100_bus": 0.6883,
1561
+ "eval_mar_100_car": 0.6094,
1562
+ "eval_mar_100_container": 0.6514,
1563
+ "eval_mar_100_motorbike": 0.4737,
1564
+ "eval_mar_large": 0.7866,
1565
+ "eval_mar_medium": 0.6078,
1566
+ "eval_mar_small": 0.3419,
1567
+ "eval_runtime": 145.0724,
1568
+ "eval_samples_per_second": 15.875,
1569
+ "eval_steps_per_second": 1.985,
1570
+ "step": 59930
1571
+ },
1572
+ {
1573
+ "epoch": 26.030368763557483,
1574
+ "grad_norm": 28658.181640625,
1575
+ "learning_rate": 6.0350030294119464e-05,
1576
+ "loss": 0.8825,
1577
+ "step": 60000
1578
+ },
1579
+ {
1580
+ "epoch": 26.247288503253795,
1581
+ "grad_norm": 6283.97509765625,
1582
+ "learning_rate": 5.979489402880387e-05,
1583
+ "loss": 0.8833,
1584
+ "step": 60500
1585
+ },
1586
+ {
1587
+ "epoch": 26.464208242950107,
1588
+ "grad_norm": 4272.13037109375,
1589
+ "learning_rate": 5.923738303304234e-05,
1590
+ "loss": 0.8844,
1591
+ "step": 61000
1592
+ },
1593
+ {
1594
+ "epoch": 26.68112798264642,
1595
+ "grad_norm": 14047.6572265625,
1596
+ "learning_rate": 5.8678680412085076e-05,
1597
+ "loss": 0.8785,
1598
+ "step": 61500
1599
+ },
1600
+ {
1601
+ "epoch": 26.89804772234273,
1602
+ "grad_norm": 8514.8544921875,
1603
+ "learning_rate": 5.811885823873863e-05,
1604
+ "loss": 0.8692,
1605
+ "step": 62000
1606
+ },
1607
+ {
1608
+ "epoch": 27.0,
1609
+ "eval_loss": 0.8606675863265991,
1610
+ "eval_map": 0.4927,
1611
+ "eval_map_50": 0.837,
1612
+ "eval_map_75": 0.5186,
1613
+ "eval_map_bus": 0.5641,
1614
+ "eval_map_car": 0.5122,
1615
+ "eval_map_container": 0.5369,
1616
+ "eval_map_large": 0.7052,
1617
+ "eval_map_medium": 0.4874,
1618
+ "eval_map_motorbike": 0.3576,
1619
+ "eval_map_small": 0.1426,
1620
+ "eval_mar_1": 0.2578,
1621
+ "eval_mar_10": 0.5566,
1622
+ "eval_mar_100": 0.6043,
1623
+ "eval_mar_100_bus": 0.6885,
1624
+ "eval_mar_100_car": 0.6043,
1625
+ "eval_mar_100_container": 0.6507,
1626
+ "eval_mar_100_motorbike": 0.4737,
1627
+ "eval_mar_large": 0.7936,
1628
+ "eval_mar_medium": 0.608,
1629
+ "eval_mar_small": 0.3251,
1630
+ "eval_runtime": 145.9029,
1631
+ "eval_samples_per_second": 15.784,
1632
+ "eval_steps_per_second": 1.974,
1633
+ "step": 62235
1634
+ },
1635
+ {
1636
+ "epoch": 27.114967462039047,
1637
+ "grad_norm": 17592.478515625,
1638
+ "learning_rate": 5.755798873023209e-05,
1639
+ "loss": 0.8719,
1640
+ "step": 62500
1641
+ },
1642
+ {
1643
+ "epoch": 27.33188720173536,
1644
+ "grad_norm": 28705.041015625,
1645
+ "learning_rate": 5.699614423890112e-05,
1646
+ "loss": 0.873,
1647
+ "step": 63000
1648
+ },
1649
+ {
1650
+ "epoch": 27.54880694143167,
1651
+ "grad_norm": 14412.8916015625,
1652
+ "learning_rate": 5.643339724285447e-05,
1653
+ "loss": 0.8654,
1654
+ "step": 63500
1655
+ },
1656
+ {
1657
+ "epoch": 27.765726681127983,
1658
+ "grad_norm": 11298.9375,
1659
+ "learning_rate": 5.587094827038153e-05,
1660
+ "loss": 0.8829,
1661
+ "step": 64000
1662
+ },
1663
+ {
1664
+ "epoch": 27.982646420824295,
1665
+ "grad_norm": 8732.7822265625,
1666
+ "learning_rate": 5.530661559735304e-05,
1667
+ "loss": 0.8694,
1668
+ "step": 64500
1669
+ },
1670
+ {
1671
+ "epoch": 28.0,
1672
+ "eval_loss": 0.8637914657592773,
1673
+ "eval_map": 0.4929,
1674
+ "eval_map_50": 0.8355,
1675
+ "eval_map_75": 0.5204,
1676
+ "eval_map_bus": 0.5621,
1677
+ "eval_map_car": 0.5152,
1678
+ "eval_map_container": 0.5343,
1679
+ "eval_map_large": 0.7119,
1680
+ "eval_map_medium": 0.4885,
1681
+ "eval_map_motorbike": 0.3601,
1682
+ "eval_map_small": 0.1308,
1683
+ "eval_mar_1": 0.2571,
1684
+ "eval_mar_10": 0.556,
1685
+ "eval_mar_100": 0.6029,
1686
+ "eval_mar_100_bus": 0.6851,
1687
+ "eval_mar_100_car": 0.6085,
1688
+ "eval_mar_100_container": 0.6431,
1689
+ "eval_mar_100_motorbike": 0.4749,
1690
+ "eval_mar_large": 0.7964,
1691
+ "eval_mar_medium": 0.6042,
1692
+ "eval_mar_small": 0.3162,
1693
+ "eval_runtime": 145.47,
1694
+ "eval_samples_per_second": 15.831,
1695
+ "eval_steps_per_second": 1.98,
1696
+ "step": 64540
1697
+ },
1698
+ {
1699
+ "epoch": 28.199566160520607,
1700
+ "grad_norm": 6033.21923828125,
1701
+ "learning_rate": 5.474159836931288e-05,
1702
+ "loss": 0.8805,
1703
+ "step": 65000
1704
+ },
1705
+ {
1706
+ "epoch": 28.41648590021692,
1707
+ "grad_norm": 7858.236328125,
1708
+ "learning_rate": 5.417596947365383e-05,
1709
+ "loss": 0.878,
1710
+ "step": 65500
1711
+ },
1712
+ {
1713
+ "epoch": 28.63340563991323,
1714
+ "grad_norm": 11765.7236328125,
1715
+ "learning_rate": 5.361093470095784e-05,
1716
+ "loss": 0.8746,
1717
+ "step": 66000
1718
+ },
1719
+ {
1720
+ "epoch": 28.850325379609544,
1721
+ "grad_norm": 3453.400146484375,
1722
+ "learning_rate": 5.304430229684755e-05,
1723
+ "loss": 0.8744,
1724
+ "step": 66500
1725
+ },
1726
+ {
1727
+ "epoch": 29.0,
1728
+ "eval_loss": 0.8675053119659424,
1729
+ "eval_map": 0.4883,
1730
+ "eval_map_50": 0.8358,
1731
+ "eval_map_75": 0.5122,
1732
+ "eval_map_bus": 0.5564,
1733
+ "eval_map_car": 0.5131,
1734
+ "eval_map_container": 0.5244,
1735
+ "eval_map_large": 0.6955,
1736
+ "eval_map_medium": 0.4818,
1737
+ "eval_map_motorbike": 0.3594,
1738
+ "eval_map_small": 0.1305,
1739
+ "eval_mar_1": 0.2542,
1740
+ "eval_mar_10": 0.5483,
1741
+ "eval_mar_100": 0.5948,
1742
+ "eval_mar_100_bus": 0.6693,
1743
+ "eval_mar_100_car": 0.602,
1744
+ "eval_mar_100_container": 0.6352,
1745
+ "eval_mar_100_motorbike": 0.4727,
1746
+ "eval_mar_large": 0.7802,
1747
+ "eval_mar_medium": 0.5936,
1748
+ "eval_mar_small": 0.3363,
1749
+ "eval_runtime": 145.2093,
1750
+ "eval_samples_per_second": 15.86,
1751
+ "eval_steps_per_second": 1.983,
1752
+ "step": 66845
1753
+ },
1754
+ {
1755
+ "epoch": 29.067245119305856,
1756
+ "grad_norm": 4124.80615234375,
1757
+ "learning_rate": 5.24772771768233e-05,
1758
+ "loss": 0.8763,
1759
+ "step": 67000
1760
+ },
1761
+ {
1762
+ "epoch": 29.284164859002168,
1763
+ "grad_norm": 12881.8603515625,
1764
+ "learning_rate": 5.1909932487296506e-05,
1765
+ "loss": 0.8783,
1766
+ "step": 67500
1767
+ },
1768
+ {
1769
+ "epoch": 29.501084598698483,
1770
+ "grad_norm": 11561.8046875,
1771
+ "learning_rate": 5.134234141590315e-05,
1772
+ "loss": 0.867,
1773
+ "step": 68000
1774
+ },
1775
+ {
1776
+ "epoch": 29.718004338394795,
1777
+ "grad_norm": 4195.20458984375,
1778
+ "learning_rate": 5.077571283466763e-05,
1779
+ "loss": 0.8623,
1780
+ "step": 68500
1781
+ },
1782
+ {
1783
+ "epoch": 29.934924078091107,
1784
+ "grad_norm": 6831.716796875,
1785
+ "learning_rate": 5.020784880687298e-05,
1786
+ "loss": 0.8608,
1787
+ "step": 69000
1788
+ },
1789
+ {
1790
+ "epoch": 30.0,
1791
+ "eval_loss": 0.8547199964523315,
1792
+ "eval_map": 0.5,
1793
+ "eval_map_50": 0.8401,
1794
+ "eval_map_75": 0.5304,
1795
+ "eval_map_bus": 0.5762,
1796
+ "eval_map_car": 0.5197,
1797
+ "eval_map_container": 0.5424,
1798
+ "eval_map_large": 0.7118,
1799
+ "eval_map_medium": 0.4962,
1800
+ "eval_map_motorbike": 0.3617,
1801
+ "eval_map_small": 0.1291,
1802
+ "eval_mar_1": 0.2596,
1803
+ "eval_mar_10": 0.559,
1804
+ "eval_mar_100": 0.6079,
1805
+ "eval_mar_100_bus": 0.6877,
1806
+ "eval_mar_100_car": 0.6139,
1807
+ "eval_mar_100_container": 0.6542,
1808
+ "eval_mar_100_motorbike": 0.4756,
1809
+ "eval_mar_large": 0.7898,
1810
+ "eval_mar_medium": 0.6093,
1811
+ "eval_mar_small": 0.34,
1812
+ "eval_runtime": 144.8308,
1813
+ "eval_samples_per_second": 15.901,
1814
+ "eval_steps_per_second": 1.989,
1815
+ "step": 69150
1816
+ },
1817
+ {
1818
+ "epoch": 30.15184381778742,
1819
+ "grad_norm": 9134.2763671875,
1820
+ "learning_rate": 4.9639957966519604e-05,
1821
+ "loss": 0.8683,
1822
+ "step": 69500
1823
+ },
1824
+ {
1825
+ "epoch": 30.36876355748373,
1826
+ "grad_norm": 6325.7216796875,
1827
+ "learning_rate": 4.907211357169706e-05,
1828
+ "loss": 0.8653,
1829
+ "step": 70000
1830
+ },
1831
+ {
1832
+ "epoch": 30.585683297180044,
1833
+ "grad_norm": 13625.853515625,
1834
+ "learning_rate": 4.850438887450347e-05,
1835
+ "loss": 0.8714,
1836
+ "step": 70500
1837
+ },
1838
+ {
1839
+ "epoch": 30.802603036876356,
1840
+ "grad_norm": 15019.505859375,
1841
+ "learning_rate": 4.793799193390669e-05,
1842
+ "loss": 0.8566,
1843
+ "step": 71000
1844
+ },
1845
+ {
1846
+ "epoch": 31.0,
1847
+ "eval_loss": 0.8513031005859375,
1848
+ "eval_map": 0.5,
1849
+ "eval_map_50": 0.8418,
1850
+ "eval_map_75": 0.5333,
1851
+ "eval_map_bus": 0.5714,
1852
+ "eval_map_car": 0.5239,
1853
+ "eval_map_container": 0.5422,
1854
+ "eval_map_large": 0.7031,
1855
+ "eval_map_medium": 0.4972,
1856
+ "eval_map_motorbike": 0.3626,
1857
+ "eval_map_small": 0.1282,
1858
+ "eval_mar_1": 0.2582,
1859
+ "eval_mar_10": 0.5594,
1860
+ "eval_mar_100": 0.6082,
1861
+ "eval_mar_100_bus": 0.6856,
1862
+ "eval_mar_100_car": 0.6153,
1863
+ "eval_mar_100_container": 0.6545,
1864
+ "eval_mar_100_motorbike": 0.4772,
1865
+ "eval_mar_large": 0.7875,
1866
+ "eval_mar_medium": 0.6116,
1867
+ "eval_mar_small": 0.3505,
1868
+ "eval_runtime": 144.5939,
1869
+ "eval_samples_per_second": 15.927,
1870
+ "eval_steps_per_second": 1.992,
1871
+ "step": 71455
1872
+ },
1873
+ {
1874
+ "epoch": 31.019522776572668,
1875
+ "grad_norm": 9826.4052734375,
1876
+ "learning_rate": 4.7370725711717407e-05,
1877
+ "loss": 0.8693,
1878
+ "step": 71500
1879
+ },
1880
+ {
1881
+ "epoch": 31.23644251626898,
1882
+ "grad_norm": 3330.38623046875,
1883
+ "learning_rate": 4.6803798666703856e-05,
1884
+ "loss": 0.8625,
1885
+ "step": 72000
1886
+ },
1887
+ {
1888
+ "epoch": 31.453362255965292,
1889
+ "grad_norm": 4638.60107421875,
1890
+ "learning_rate": 4.623728393262576e-05,
1891
+ "loss": 0.8672,
1892
+ "step": 72500
1893
+ },
1894
+ {
1895
+ "epoch": 31.670281995661604,
1896
+ "grad_norm": 10907.923828125,
1897
+ "learning_rate": 4.5672386115774904e-05,
1898
+ "loss": 0.8625,
1899
+ "step": 73000
1900
+ },
1901
+ {
1902
+ "epoch": 31.887201735357916,
1903
+ "grad_norm": 3609.915283203125,
1904
+ "learning_rate": 4.5106913993021883e-05,
1905
+ "loss": 0.8586,
1906
+ "step": 73500
1907
+ },
1908
+ {
1909
+ "epoch": 32.0,
1910
+ "eval_loss": 0.8532082438468933,
1911
+ "eval_map": 0.5035,
1912
+ "eval_map_50": 0.8414,
1913
+ "eval_map_75": 0.538,
1914
+ "eval_map_bus": 0.5799,
1915
+ "eval_map_car": 0.5235,
1916
+ "eval_map_container": 0.5481,
1917
+ "eval_map_large": 0.7084,
1918
+ "eval_map_medium": 0.4993,
1919
+ "eval_map_motorbike": 0.3626,
1920
+ "eval_map_small": 0.1292,
1921
+ "eval_mar_1": 0.2603,
1922
+ "eval_mar_10": 0.5621,
1923
+ "eval_mar_100": 0.6094,
1924
+ "eval_mar_100_bus": 0.6893,
1925
+ "eval_mar_100_car": 0.6139,
1926
+ "eval_mar_100_container": 0.6587,
1927
+ "eval_mar_100_motorbike": 0.4758,
1928
+ "eval_mar_large": 0.7924,
1929
+ "eval_mar_medium": 0.6135,
1930
+ "eval_mar_small": 0.3286,
1931
+ "eval_runtime": 145.9881,
1932
+ "eval_samples_per_second": 15.775,
1933
+ "eval_steps_per_second": 1.973,
1934
+ "step": 73760
1935
+ },
1936
+ {
1937
+ "epoch": 32.10412147505423,
1938
+ "grad_norm": 11651.8828125,
1939
+ "learning_rate": 4.4542073079838684e-05,
1940
+ "loss": 0.8618,
1941
+ "step": 74000
1942
+ },
1943
+ {
1944
+ "epoch": 32.321041214750544,
1945
+ "grad_norm": 17436.388671875,
1946
+ "learning_rate": 4.39779362408734e-05,
1947
+ "loss": 0.8542,
1948
+ "step": 74500
1949
+ },
1950
+ {
1951
+ "epoch": 32.53796095444685,
1952
+ "grad_norm": 20459.701171875,
1953
+ "learning_rate": 4.341457624994828e-05,
1954
+ "loss": 0.8554,
1955
+ "step": 75000
1956
+ },
1957
+ {
1958
+ "epoch": 32.75488069414317,
1959
+ "grad_norm": 5393.4375,
1960
+ "learning_rate": 4.285318990553778e-05,
1961
+ "loss": 0.8528,
1962
+ "step": 75500
1963
+ },
1964
+ {
1965
+ "epoch": 32.971800433839476,
1966
+ "grad_norm": 4986.16796875,
1967
+ "learning_rate": 4.229159960541767e-05,
1968
+ "loss": 0.8598,
1969
+ "step": 76000
1970
+ },
1971
+ {
1972
+ "epoch": 33.0,
1973
+ "eval_loss": 0.8493744134902954,
1974
+ "eval_map": 0.499,
1975
+ "eval_map_50": 0.8415,
1976
+ "eval_map_75": 0.5323,
1977
+ "eval_map_bus": 0.5708,
1978
+ "eval_map_car": 0.5215,
1979
+ "eval_map_container": 0.5432,
1980
+ "eval_map_large": 0.7104,
1981
+ "eval_map_medium": 0.4969,
1982
+ "eval_map_motorbike": 0.3606,
1983
+ "eval_map_small": 0.1301,
1984
+ "eval_mar_1": 0.2582,
1985
+ "eval_mar_10": 0.5605,
1986
+ "eval_mar_100": 0.6084,
1987
+ "eval_mar_100_bus": 0.6857,
1988
+ "eval_mar_100_car": 0.615,
1989
+ "eval_mar_100_container": 0.6551,
1990
+ "eval_mar_100_motorbike": 0.4778,
1991
+ "eval_mar_large": 0.7942,
1992
+ "eval_mar_medium": 0.6113,
1993
+ "eval_mar_small": 0.3277,
1994
+ "eval_runtime": 144.7755,
1995
+ "eval_samples_per_second": 15.907,
1996
+ "eval_steps_per_second": 1.989,
1997
+ "step": 76065
1998
+ },
1999
+ {
2000
+ "epoch": 33.18872017353579,
2001
+ "grad_norm": 6646.677734375,
2002
+ "learning_rate": 4.173212384261402e-05,
2003
+ "loss": 0.8671,
2004
+ "step": 76500
2005
+ },
2006
+ {
2007
+ "epoch": 33.40563991323211,
2008
+ "grad_norm": 4728.1552734375,
2009
+ "learning_rate": 4.1172592425996225e-05,
2010
+ "loss": 0.8671,
2011
+ "step": 77000
2012
+ },
2013
+ {
2014
+ "epoch": 33.622559652928416,
2015
+ "grad_norm": 9825.3720703125,
2016
+ "learning_rate": 4.0614199747616364e-05,
2017
+ "loss": 0.8668,
2018
+ "step": 77500
2019
+ },
2020
+ {
2021
+ "epoch": 33.83947939262473,
2022
+ "grad_norm": 3324.343505859375,
2023
+ "learning_rate": 4.0057017840298235e-05,
2024
+ "loss": 0.8728,
2025
+ "step": 78000
2026
+ },
2027
+ {
2028
+ "epoch": 34.0,
2029
+ "eval_loss": 0.8495970368385315,
2030
+ "eval_map": 0.5027,
2031
+ "eval_map_50": 0.841,
2032
+ "eval_map_75": 0.5357,
2033
+ "eval_map_bus": 0.5809,
2034
+ "eval_map_car": 0.5196,
2035
+ "eval_map_container": 0.5452,
2036
+ "eval_map_large": 0.715,
2037
+ "eval_map_medium": 0.4969,
2038
+ "eval_map_motorbike": 0.3652,
2039
+ "eval_map_small": 0.1332,
2040
+ "eval_mar_1": 0.2609,
2041
+ "eval_mar_10": 0.5617,
2042
+ "eval_mar_100": 0.6105,
2043
+ "eval_mar_100_bus": 0.6941,
2044
+ "eval_mar_100_car": 0.6116,
2045
+ "eval_mar_100_container": 0.6565,
2046
+ "eval_mar_100_motorbike": 0.4799,
2047
+ "eval_mar_large": 0.7994,
2048
+ "eval_mar_medium": 0.6147,
2049
+ "eval_mar_small": 0.3335,
2050
+ "eval_runtime": 145.9053,
2051
+ "eval_samples_per_second": 15.784,
2052
+ "eval_steps_per_second": 1.974,
2053
+ "step": 78370
2054
+ },
2055
+ {
2056
+ "epoch": 34.05639913232104,
2057
+ "grad_norm": 20089.041015625,
2058
+ "learning_rate": 3.950111858067586e-05,
2059
+ "loss": 0.8606,
2060
+ "step": 78500
2061
+ },
2062
+ {
2063
+ "epoch": 34.273318872017356,
2064
+ "grad_norm": 22944.431640625,
2065
+ "learning_rate": 3.894657367992129e-05,
2066
+ "loss": 0.8592,
2067
+ "step": 79000
2068
+ },
2069
+ {
2070
+ "epoch": 34.490238611713664,
2071
+ "grad_norm": 6247.44970703125,
2072
+ "learning_rate": 3.8393454674493866e-05,
2073
+ "loss": 0.8535,
2074
+ "step": 79500
2075
+ },
2076
+ {
2077
+ "epoch": 34.70715835140998,
2078
+ "grad_norm": 8178.81982421875,
2079
+ "learning_rate": 3.7841832916912035e-05,
2080
+ "loss": 0.8499,
2081
+ "step": 80000
2082
+ },
2083
+ {
2084
+ "epoch": 34.92407809110629,
2085
+ "grad_norm": 6138.80712890625,
2086
+ "learning_rate": 3.7291779566548805e-05,
2087
+ "loss": 0.8496,
2088
+ "step": 80500
2089
+ },
2090
+ {
2091
+ "epoch": 35.0,
2092
+ "eval_loss": 0.8385424613952637,
2093
+ "eval_map": 0.5081,
2094
+ "eval_map_50": 0.8476,
2095
+ "eval_map_75": 0.5424,
2096
+ "eval_map_bus": 0.5828,
2097
+ "eval_map_car": 0.5317,
2098
+ "eval_map_container": 0.5475,
2099
+ "eval_map_large": 0.7206,
2100
+ "eval_map_medium": 0.504,
2101
+ "eval_map_motorbike": 0.3703,
2102
+ "eval_map_small": 0.1423,
2103
+ "eval_mar_1": 0.2623,
2104
+ "eval_mar_10": 0.5665,
2105
+ "eval_mar_100": 0.6148,
2106
+ "eval_mar_100_bus": 0.6954,
2107
+ "eval_mar_100_car": 0.6213,
2108
+ "eval_mar_100_container": 0.6591,
2109
+ "eval_mar_100_motorbike": 0.4833,
2110
+ "eval_mar_large": 0.8048,
2111
+ "eval_mar_medium": 0.6188,
2112
+ "eval_mar_small": 0.3475,
2113
+ "eval_runtime": 144.7118,
2114
+ "eval_samples_per_second": 15.914,
2115
+ "eval_steps_per_second": 1.99,
2116
+ "step": 80675
2117
+ },
2118
+ {
2119
+ "epoch": 35.140997830802604,
2120
+ "grad_norm": 5527.109375,
2121
+ "learning_rate": 3.674446072528038e-05,
2122
+ "loss": 0.8416,
2123
+ "step": 81000
2124
+ },
2125
+ {
2126
+ "epoch": 35.35791757049891,
2127
+ "grad_norm": 6165.18603515625,
2128
+ "learning_rate": 3.619775335834121e-05,
2129
+ "loss": 0.8443,
2130
+ "step": 81500
2131
+ },
2132
+ {
2133
+ "epoch": 35.57483731019523,
2134
+ "grad_norm": 7007.72900390625,
2135
+ "learning_rate": 3.5652826485379264e-05,
2136
+ "loss": 0.8546,
2137
+ "step": 82000
2138
+ },
2139
+ {
2140
+ "epoch": 35.79175704989154,
2141
+ "grad_norm": 7134.87548828125,
2142
+ "learning_rate": 3.5109750402125426e-05,
2143
+ "loss": 0.8477,
2144
+ "step": 82500
2145
+ },
2146
+ {
2147
+ "epoch": 36.0,
2148
+ "eval_loss": 0.840656042098999,
2149
+ "eval_map": 0.5078,
2150
+ "eval_map_50": 0.8475,
2151
+ "eval_map_75": 0.543,
2152
+ "eval_map_bus": 0.5808,
2153
+ "eval_map_car": 0.5308,
2154
+ "eval_map_container": 0.5506,
2155
+ "eval_map_large": 0.7143,
2156
+ "eval_map_medium": 0.5038,
2157
+ "eval_map_motorbike": 0.3689,
2158
+ "eval_map_small": 0.1375,
2159
+ "eval_mar_1": 0.2612,
2160
+ "eval_mar_10": 0.5658,
2161
+ "eval_mar_100": 0.6127,
2162
+ "eval_mar_100_bus": 0.6911,
2163
+ "eval_mar_100_car": 0.6198,
2164
+ "eval_mar_100_container": 0.6566,
2165
+ "eval_mar_100_motorbike": 0.4835,
2166
+ "eval_mar_large": 0.7981,
2167
+ "eval_mar_medium": 0.6156,
2168
+ "eval_mar_small": 0.3327,
2169
+ "eval_runtime": 144.8577,
2170
+ "eval_samples_per_second": 15.898,
2171
+ "eval_steps_per_second": 1.988,
2172
+ "step": 82980
2173
+ }
2174
+ ],
2175
+ "logging_steps": 500,
2176
+ "max_steps": 138300,
2177
+ "num_input_tokens_seen": 0,
2178
+ "num_train_epochs": 60,
2179
+ "save_steps": 500,
2180
+ "stateful_callbacks": {
2181
+ "TrainerControl": {
2182
+ "args": {
2183
+ "should_epoch_stop": false,
2184
+ "should_evaluate": false,
2185
+ "should_log": false,
2186
+ "should_save": true,
2187
+ "should_training_stop": false
2188
+ },
2189
+ "attributes": {}
2190
+ }
2191
+ },
2192
+ "total_flos": 1.526046904086461e+20,
2193
+ "train_batch_size": 4,
2194
+ "trial_name": null,
2195
+ "trial_params": null
2196
+ }
checkpoint-82980/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e79ce30220cc0f492cce991ef1b82af549d14310ccde1c3ab2fa0987f62e3936
3
+ size 5240