Commit
·
d7b681a
1
Parent(s):
f5b09b7
Upload cp_dermamnist model (LFS)
Browse files
cp_dermamnist/ckpt/final/__0_0.distcp
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:59c7396c338cbee0988cfcd2dd7776eaf835f5b271dac9f70f16d8fca2129a24
|
| 3 |
+
size 2544882430
|
cp_dermamnist/config.yaml
ADDED
|
@@ -0,0 +1,188 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
MODEL:
|
| 2 |
+
META_ARCHITECTURE: SSLMetaArch
|
| 3 |
+
DEVICE: cuda
|
| 4 |
+
WEIGHTS: ''
|
| 5 |
+
DTYPE: float32
|
| 6 |
+
compute_precision:
|
| 7 |
+
param_dtype: bf16
|
| 8 |
+
reduce_dtype: fp32
|
| 9 |
+
sharding_strategy: SHARD_GRAD_OP
|
| 10 |
+
dino:
|
| 11 |
+
loss_weight: 1.0
|
| 12 |
+
global_ignore_diagonal: true
|
| 13 |
+
head_n_prototypes: 65536
|
| 14 |
+
head_bottleneck_dim: 256
|
| 15 |
+
head_norm_last_layer: false
|
| 16 |
+
head_nlayers: 3
|
| 17 |
+
head_hidden_dim: 1536
|
| 18 |
+
koleo_loss_weight: 0.1
|
| 19 |
+
koleo_loss_distributed: false
|
| 20 |
+
koleo_topk: 1
|
| 21 |
+
koleo_distributed_replicas: 0
|
| 22 |
+
koleo_distributed_loss_group_size: null
|
| 23 |
+
koleo_distributed_loss_group_data: true
|
| 24 |
+
force_weight_norm: false
|
| 25 |
+
reweight_dino_local_loss: false
|
| 26 |
+
local_loss_weight_schedule:
|
| 27 |
+
start: 0.5
|
| 28 |
+
peak: 0.5
|
| 29 |
+
end: 0.5
|
| 30 |
+
warmup_epochs: 0
|
| 31 |
+
ibot:
|
| 32 |
+
loss_weight: 1.0
|
| 33 |
+
mask_sample_probability: 0.5
|
| 34 |
+
mask_ratio_min_max:
|
| 35 |
+
- 0.1
|
| 36 |
+
- 0.5
|
| 37 |
+
mask_random_circular_shift: false
|
| 38 |
+
force_masking_even_with_zero_weight: false
|
| 39 |
+
separate_head: true
|
| 40 |
+
head_n_prototypes: 65536
|
| 41 |
+
head_bottleneck_dim: 256
|
| 42 |
+
head_norm_last_layer: false
|
| 43 |
+
head_nlayers: 3
|
| 44 |
+
head_hidden_dim: 1536
|
| 45 |
+
gram:
|
| 46 |
+
use_loss: false
|
| 47 |
+
compute_stats: false
|
| 48 |
+
loss_weight: 1.0
|
| 49 |
+
ema_teacher: false
|
| 50 |
+
ckpt: null
|
| 51 |
+
it_load_ema_teacher: -1
|
| 52 |
+
rep_update: true
|
| 53 |
+
update_frequency: 50000
|
| 54 |
+
it_first_update: 0
|
| 55 |
+
max_updates: null
|
| 56 |
+
normalized: true
|
| 57 |
+
img_level: false
|
| 58 |
+
remove_neg: false
|
| 59 |
+
remove_only_teacher_neg: false
|
| 60 |
+
tokens_used: all
|
| 61 |
+
global_teacher_resize_method: bicubic
|
| 62 |
+
global_teacher_resize_antialias: false
|
| 63 |
+
loss_weight_schedule: null
|
| 64 |
+
train:
|
| 65 |
+
batch_size_per_gpu: 64
|
| 66 |
+
dataset_path: dermamnist:root=/root/data/medmnist:split=TRAIN:limit_data=10
|
| 67 |
+
data_config: null
|
| 68 |
+
output_dir: /root/output/10/cp_dermamnist
|
| 69 |
+
saveckp_freq: 20
|
| 70 |
+
seed: 0
|
| 71 |
+
num_workers: 10
|
| 72 |
+
OFFICIAL_EPOCH_LENGTH: 16
|
| 73 |
+
monitor_gradient_norm: false
|
| 74 |
+
chunk_schedule: []
|
| 75 |
+
use_teacher_head: true
|
| 76 |
+
learn_from_teacher_tokens: false
|
| 77 |
+
centering: sinkhorn_knopp
|
| 78 |
+
checkpointing: false
|
| 79 |
+
checkpointing_full: false
|
| 80 |
+
compile: true
|
| 81 |
+
cudagraphs: false
|
| 82 |
+
sharded_eval_checkpoint: false
|
| 83 |
+
cache_dataset: true
|
| 84 |
+
cell_augmentation: false
|
| 85 |
+
cell_augmentation_type: hpa
|
| 86 |
+
student:
|
| 87 |
+
arch: vit_base
|
| 88 |
+
patch_size: 16
|
| 89 |
+
drop_path_rate: 0.3
|
| 90 |
+
layerscale: 1.0e-05
|
| 91 |
+
pretrained_weights: ''
|
| 92 |
+
ffn_layer: mlp
|
| 93 |
+
ffn_ratio: 4.0
|
| 94 |
+
resume_from_teacher_chkpt: ''
|
| 95 |
+
qkv_bias: true
|
| 96 |
+
proj_bias: true
|
| 97 |
+
ffn_bias: true
|
| 98 |
+
norm_layer: layernorm
|
| 99 |
+
n_storage_tokens: 4
|
| 100 |
+
mask_k_bias: false
|
| 101 |
+
untie_cls_and_patch_norms: false
|
| 102 |
+
untie_global_and_local_cls_norm: false
|
| 103 |
+
in_chans: 3
|
| 104 |
+
pos_embed_type: rope
|
| 105 |
+
pos_embed_rope_base: 100.0
|
| 106 |
+
pos_embed_rope_min_period: null
|
| 107 |
+
pos_embed_rope_max_period: null
|
| 108 |
+
pos_embed_rope_normalize_coords: separate
|
| 109 |
+
pos_embed_rope_shift_coords: null
|
| 110 |
+
pos_embed_rope_jitter_coords: null
|
| 111 |
+
pos_embed_rope_rescale_coords: null
|
| 112 |
+
pos_embed_rope_dtype: bf16
|
| 113 |
+
fp8_enabled: false
|
| 114 |
+
fp8_filter: blocks
|
| 115 |
+
patch_drop: 0.0
|
| 116 |
+
teacher:
|
| 117 |
+
momentum_teacher: 0.992
|
| 118 |
+
final_momentum_teacher: 1
|
| 119 |
+
warmup_teacher_temp: 0.04
|
| 120 |
+
teacher_temp: 0.07
|
| 121 |
+
warmup_teacher_temp_epochs: 30
|
| 122 |
+
in_chans: 3
|
| 123 |
+
distillation:
|
| 124 |
+
enabled: false
|
| 125 |
+
full_cfg_path: ''
|
| 126 |
+
checkpoint_path: ''
|
| 127 |
+
multidistillation:
|
| 128 |
+
enabled: false
|
| 129 |
+
cp:
|
| 130 |
+
enabled: true
|
| 131 |
+
hf_model: facebook/dinov3-vitb16-pretrain-lvd1689m
|
| 132 |
+
checkpoint_url: ''
|
| 133 |
+
hrft:
|
| 134 |
+
enabled: false
|
| 135 |
+
checkpoint_path: ''
|
| 136 |
+
optim:
|
| 137 |
+
epochs: 100
|
| 138 |
+
optimizer: adamw
|
| 139 |
+
weight_decay: 0.04
|
| 140 |
+
weight_decay_end: 0.4
|
| 141 |
+
lr: 0.001
|
| 142 |
+
warmup_epochs: 10
|
| 143 |
+
min_lr: 1.0e-06
|
| 144 |
+
schedule_trunc_extra: 0.0
|
| 145 |
+
clip_grad: 3.0
|
| 146 |
+
freeze_last_layer_epochs: 1
|
| 147 |
+
scaling_rule: sqrt_wrt_1024
|
| 148 |
+
patch_embed_lr_mult: 0.2
|
| 149 |
+
dino_head_wd_multiplier: 1.0
|
| 150 |
+
layerwise_decay: 0.9
|
| 151 |
+
multi_tensor_optim: true
|
| 152 |
+
dump_fsdp_weights_path: ''
|
| 153 |
+
adamw_beta1: 0.9
|
| 154 |
+
adamw_beta2: 0.999
|
| 155 |
+
crops:
|
| 156 |
+
global_crops_scale:
|
| 157 |
+
- 0.32
|
| 158 |
+
- 1.0
|
| 159 |
+
local_crops_number: 8
|
| 160 |
+
local_crops_scale:
|
| 161 |
+
- 0.05
|
| 162 |
+
- 0.32
|
| 163 |
+
global_crops_size: 224
|
| 164 |
+
local_crops_size: 96
|
| 165 |
+
global_local_crop_pairs_ratios: 1.0
|
| 166 |
+
gram_teacher_crops_size: null
|
| 167 |
+
localcrops_subset_of_globalcrops: false
|
| 168 |
+
share_color_jitter: false
|
| 169 |
+
horizontal_flips: true
|
| 170 |
+
gram_teacher_no_distortions: false
|
| 171 |
+
rgb_mean:
|
| 172 |
+
- 0.4846
|
| 173 |
+
- 0.4846
|
| 174 |
+
- 0.4846
|
| 175 |
+
rgb_std:
|
| 176 |
+
- 0.2522
|
| 177 |
+
- 0.2522
|
| 178 |
+
- 0.2522
|
| 179 |
+
evaluation:
|
| 180 |
+
eval_period_iterations: 12500
|
| 181 |
+
low_freq_every: 5
|
| 182 |
+
config_files:
|
| 183 |
+
high_freq: benchmark_high_frequency.yaml
|
| 184 |
+
low_freq: benchmark_low_frequency.yaml
|
| 185 |
+
checkpointing:
|
| 186 |
+
period: 3750
|
| 187 |
+
max_to_keep: 3
|
| 188 |
+
keep_every: 99999999999999999
|
cp_dermamnist/logs/log.txt
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
cp_dermamnist/nan_logs/logs/log.txt
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
D20251218 11:17:59 51354 nan_logger __init__.py:132] PyTorch distributed environment: manual job using 127.0.0.1:42435 (rank=0, world size=1)
|
cp_dermamnist/training_metrics.json
ADDED
|
@@ -0,0 +1,161 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"iteration": 0, "iter_time": 6.444757461547852, "data_time": 2.76727032661438, "lr": 0.0, "wd": 0.03999999999999998, "mom": 0.992, "last_layer_lr": 0.0, "total_loss": 16.668537139892578, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 11.109793663024902, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 11.109075546264648, "koleo_loss": 0.018310546875, "ibot_loss": 5.555161476135254, "backbone_grad_norm": 1.8635529279708862, "dino_head_grad_norm": 0.014560514129698277, "ibot_head_grad_norm": 0.017493214458227158}
|
| 2 |
+
{"iteration": 10, "iter_time": 0.7895175814628601, "data_time": 0.2521049380302429, "lr": 3.144654088050314e-05, "wd": 0.04000867438732092, "mom": 0.9920001927641626, "last_layer_lr": 0.0, "total_loss": 16.66341781616211, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 11.109664916992188, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 11.107802391052246, "koleo_loss": -0.0047607421875, "ibot_loss": 5.554594039916992, "backbone_grad_norm": 1.4138129949569702, "dino_head_grad_norm": 0.014028672128915787, "ibot_head_grad_norm": 0.01770169660449028}
|
| 3 |
+
{"iteration": 20, "iter_time": 0.22447450459003448, "data_time": 0.0006041526794433594, "lr": 6.289308176100629e-05, "wd": 0.04003469671322829, "mom": 0.9920007710380717, "last_layer_lr": 0.0, "total_loss": 16.647686004638672, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 11.108844757080078, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 11.107467651367188, "koleo_loss": -0.076171875, "ibot_loss": 5.554446697235107, "backbone_grad_norm": 1.3468706607818604, "dino_head_grad_norm": 0.013870468363165855, "ibot_head_grad_norm": 0.01770169660449028}
|
| 4 |
+
{"iteration": 30, "iter_time": 0.22879746556282043, "data_time": 0.0008204460027627647, "lr": 0.00012578616352201257, "wd": 0.04013877347666983, "mom": 0.9920030838550371, "last_layer_lr": 0.00012578616352201257, "total_loss": 16.62700653076172, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 11.104808807373047, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 11.10318374633789, "koleo_loss": -0.1494140625, "ibot_loss": 5.553459167480469, "backbone_grad_norm": 1.1784331798553467, "dino_head_grad_norm": 0.01424410566687584, "ibot_head_grad_norm": 0.01718372292816639}
|
| 5 |
+
{"iteration": 40, "iter_time": 0.23259754478931427, "data_time": 0.000977289630100131, "lr": 0.00018867924528301889, "wd": 0.04031219016675108, "mom": 0.9920069375592612, "last_layer_lr": 0.00018867924528301889, "total_loss": 16.601362228393555, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 11.09149169921875, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 11.09079360961914, "koleo_loss": -0.18359375, "ibot_loss": 5.54946231842041, "backbone_grad_norm": 0.992458701133728, "dino_head_grad_norm": 0.010159310884773731, "ibot_head_grad_norm": 0.014198572374880314}
|
| 6 |
+
{"iteration": 50, "iter_time": 0.2886291444301605, "data_time": 0.056966375559568405, "lr": 0.00025157232704402514, "wd": 0.04055487992803691, "mom": 0.9920123306650674, "last_layer_lr": 0.00025157232704402514, "total_loss": 16.591554641723633, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 11.087812423706055, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 11.087587356567383, "koleo_loss": -0.2109375, "ibot_loss": 5.545874118804932, "backbone_grad_norm": 0.9951783418655396, "dino_head_grad_norm": 0.005757933482527733, "ibot_head_grad_norm": 0.0053735594265162945}
|
| 7 |
+
{"iteration": 60, "iter_time": 0.3120535910129547, "data_time": 0.08007480204105377, "lr": 0.0003144654088050315, "wd": 0.040866749199004515, "mom": 0.9920192610933112, "last_layer_lr": 0.0003144654088050315, "total_loss": 16.583526611328125, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 11.085367202758789, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 11.085355758666992, "koleo_loss": -0.236328125, "ibot_loss": 5.545224666595459, "backbone_grad_norm": 0.9767703413963318, "dino_head_grad_norm": 0.006228955928236246, "ibot_head_grad_norm": 0.00309204775840044}
|
| 8 |
+
{"iteration": 70, "iter_time": 0.30343711376190186, "data_time": 0.07256388664245605, "lr": 0.00037735849056603777, "wd": 0.04124767774811322, "mom": 0.9920277261721803, "last_layer_lr": 0.00037735849056603777, "total_loss": 16.57775115966797, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 11.081573486328125, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 11.083568572998047, "koleo_loss": -0.2451171875, "ibot_loss": 5.545007705688477, "backbone_grad_norm": 0.8772826790809631, "dino_head_grad_norm": 0.007231824100017548, "ibot_head_grad_norm": 0.002225105185061693}
|
| 9 |
+
{"iteration": 80, "iter_time": 0.33042094111442566, "data_time": 0.10071263462305069, "lr": 0.00044025157232704406, "wd": 0.04169751872015609, "mom": 0.9920377226382256, "last_layer_lr": 0.00044025157232704406, "total_loss": 16.571788787841797, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 11.077142715454102, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 11.079001426696777, "koleo_loss": -0.248046875, "ibot_loss": 5.544842720031738, "backbone_grad_norm": 0.7543807625770569, "dino_head_grad_norm": 0.008663348853588104, "ibot_head_grad_norm": 0.0020751412957906723}
|
| 10 |
+
{"iteration": 90, "iter_time": 0.32583174109458923, "data_time": 0.09686288982629776, "lr": 0.0005031446540880503, "wd": 0.04221609869287518, "mom": 0.9920492466376194, "last_layer_lr": 0.0005031446540880503, "total_loss": 16.566713333129883, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 11.072535514831543, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 11.075629234313965, "koleo_loss": -0.248046875, "ibot_loss": 5.544731616973877, "backbone_grad_norm": 0.9868071675300598, "dino_head_grad_norm": 0.009687134064733982, "ibot_head_grad_norm": 0.002093437360599637}
|
| 11 |
+
{"iteration": 100, "iter_time": 0.31241199374198914, "data_time": 0.08449149131774902, "lr": 0.0005660377358490566, "wd": 0.04280321774381951, "mom": 0.9920622937276404, "last_layer_lr": 0.0005660377358490566, "total_loss": 16.5623779296875, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 11.066414833068848, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 11.069883346557617, "koleo_loss": -0.25, "ibot_loss": 5.5445427894592285, "backbone_grad_norm": 0.9868071675300598, "dino_head_grad_norm": 0.009597714990377426, "ibot_head_grad_norm": 0.002289612777531147}
|
| 12 |
+
{"iteration": 110, "iter_time": 0.27893462777137756, "data_time": 0.0505380742251873, "lr": 0.000628930817610063, "wd": 0.043458649527418514, "mom": 0.992076858878387, "last_layer_lr": 0.000628930817610063, "total_loss": 16.555429458618164, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 11.05999755859375, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 11.063634872436523, "koleo_loss": -0.2470703125, "ibot_loss": 5.544259548187256, "backbone_grad_norm": 0.7889320850372314, "dino_head_grad_norm": 0.009755908511579037, "ibot_head_grad_norm": 0.0024377575609833}
|
| 13 |
+
{"iteration": 120, "iter_time": 0.2450180947780609, "data_time": 0.014536917209625244, "lr": 0.0006918238993710692, "wd": 0.044182141362242966, "mom": 0.9920929364747165, "last_layer_lr": 0.0006918238993710692, "total_loss": 16.54768943786621, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 11.053613662719727, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 11.05710506439209, "koleo_loss": -0.24609375, "ibot_loss": 5.543905735015869, "backbone_grad_norm": 0.6488502025604248, "dino_head_grad_norm": 0.010497678071260452, "ibot_head_grad_norm": 0.002566511509940028}
|
| 14 |
+
{"iteration": 130, "iter_time": 0.24050787091255188, "data_time": 0.008593070320785046, "lr": 0.0007547169811320755, "wd": 0.044973414328418226, "mom": 0.9921105203184093, "last_layer_lr": 0.0007547169811320755, "total_loss": 16.541244506835938, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 11.045965194702148, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 11.050857543945312, "koleo_loss": -0.248046875, "ibot_loss": 5.543336868286133, "backbone_grad_norm": 0.5542994737625122, "dino_head_grad_norm": 0.010324962437152863, "ibot_head_grad_norm": 0.002727126469835639}
|
| 15 |
+
{"iteration": 140, "iter_time": 0.23874135315418243, "data_time": 0.006759464740753174, "lr": 0.0008176100628930818, "wd": 0.04583216337515322, "mom": 0.9921296036305589, "last_layer_lr": 0.0008176100628930818, "total_loss": 16.532733917236328, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 11.038932800292969, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 11.041606903076172, "koleo_loss": -0.248046875, "ibot_loss": 5.542904853820801, "backbone_grad_norm": 0.49086758494377136, "dino_head_grad_norm": 0.010324962437152863, "ibot_head_grad_norm": 0.002782830037176609}
|
| 16 |
+
{"iteration": 150, "iter_time": 0.2786220908164978, "data_time": 0.029589582234621048, "lr": 0.0008805031446540881, "wd": 0.04675805743834349, "mom": 0.9921501790541855, "last_layer_lr": 0.0008805031446540881, "total_loss": 16.525047302246094, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 11.030937194824219, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 11.03408145904541, "koleo_loss": -0.24609375, "ibot_loss": 5.542627334594727, "backbone_grad_norm": 0.4651944935321808, "dino_head_grad_norm": 0.0127640962600708, "ibot_head_grad_norm": 0.0030178148299455643}
|
| 17 |
+
{"iteration": 160, "iter_time": 0.29093027114868164, "data_time": 0.044675588607788086, "lr": 0.0009433962264150943, "wd": 0.047750739568202416, "mom": 0.9921722386570712, "last_layer_lr": 0.0009433962264150943, "total_loss": 16.51519012451172, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 11.021543502807617, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 11.025308609008789, "koleo_loss": -0.244140625, "ibot_loss": 5.541887283325195, "backbone_grad_norm": 0.39948493242263794, "dino_head_grad_norm": 0.013525830581784248, "ibot_head_grad_norm": 0.002988224383443594}
|
| 18 |
+
{"iteration": 170, "iter_time": 0.292701780796051, "data_time": 0.06427943706512451, "lr": 0.0009999037166207915, "wd": 0.04880982706687231, "mom": 0.9921957739348194, "last_layer_lr": 0.0009999037166207915, "total_loss": 16.50364112854004, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 11.010187149047852, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 11.018043518066406, "koleo_loss": -0.2431640625, "ibot_loss": 5.541452407836914, "backbone_grad_norm": 0.3601989448070526, "dino_head_grad_norm": 0.013371377252042294, "ibot_head_grad_norm": 0.0031630895100533962}
|
| 19 |
+
{"iteration": 180, "iter_time": 0.3246801495552063, "data_time": 0.09362369775772095, "lr": 0.0009998561715549793, "wd": 0.04993491163596192, "mom": 0.9922207758141325, "last_layer_lr": 0.0009998561715549793, "total_loss": 16.49448585510254, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 11.000787734985352, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 11.011016845703125, "koleo_loss": -0.240234375, "ibot_loss": 5.5408406257629395, "backbone_grad_norm": 0.3762800693511963, "dino_head_grad_norm": 0.012780810706317425, "ibot_head_grad_norm": 0.0035500137601047754}
|
| 20 |
+
{"iteration": 190, "iter_time": 0.2911168336868286, "data_time": 0.06070924922823906, "lr": 0.0009994758653387506, "wd": 0.051125559533952813, "mom": 0.99224723465631, "last_layer_lr": 0.0009994758653387506, "total_loss": 16.482898712158203, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.989625930786133, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.9995698928833, "koleo_loss": -0.240234375, "ibot_loss": 5.53983211517334, "backbone_grad_norm": 0.3582225739955902, "dino_head_grad_norm": 0.012780810706317425, "ibot_head_grad_norm": 0.00323619251139462}
|
| 21 |
+
{"iteration": 200, "iter_time": 0.2463904172182083, "data_time": 0.019059956073760986, "lr": 0.000998858073635063, "wd": 0.05238131174341537, "mom": 0.9922751402609647, "last_layer_lr": 0.000998858073635063, "total_loss": 16.468568801879883, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.976191520690918, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.987922668457031, "koleo_loss": -0.2431640625, "ibot_loss": 5.539295673370361, "backbone_grad_norm": 0.3902455270290375, "dino_head_grad_norm": 0.013988278806209564, "ibot_head_grad_norm": 0.0030467892065644264}
|
| 22 |
+
{"iteration": 210, "iter_time": 0.2843726575374603, "data_time": 0.05567450448870659, "lr": 0.0009980030904793076, "wd": 0.05370168414796839, "mom": 0.9923044818699549, "last_layer_lr": 0.0009980030904793076, "total_loss": 16.455322265625, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.963926315307617, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.973917961120605, "koleo_loss": -0.244140625, "ibot_loss": 5.537728309631348, "backbone_grad_norm": 0.40769466757774353, "dino_head_grad_norm": 0.012832478620111942, "ibot_head_grad_norm": 0.003031024942174554}
|
| 23 |
+
{"iteration": 220, "iter_time": 0.2836276888847351, "data_time": 0.051857758313417435, "lr": 0.0009969113227971608, "wd": 0.0550861677189155, "mom": 0.9923352481715314, "last_layer_lr": 0.0009969113227971608, "total_loss": 16.44082260131836, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.95071029663086, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.960681915283203, "koleo_loss": -0.240234375, "ibot_loss": 5.536869049072266, "backbone_grad_norm": 0.3724404573440552, "dino_head_grad_norm": 0.013739281333982944, "ibot_head_grad_norm": 0.0034543806686997414}
|
| 24 |
+
{"iteration": 230, "iter_time": 0.24833670258522034, "data_time": 0.01616353914141655, "lr": 0.00099558329021091, "wd": 0.056534228711485335, "mom": 0.9923674273046996, "last_layer_lr": 0.00099558329021091, "total_loss": 16.425334930419922, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.938187599182129, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.950660705566406, "koleo_loss": -0.2392578125, "ibot_loss": 5.535258769989014, "backbone_grad_norm": 0.3785460293292999, "dino_head_grad_norm": 0.01587742380797863, "ibot_head_grad_norm": 0.003849756671115756}
|
| 25 |
+
{"iteration": 240, "iter_time": 0.297709196805954, "data_time": 0.0651877373456955, "lr": 0.0009940196247921393, "wd": 0.05804530887060011, "mom": 0.9924010068637911, "last_layer_lr": 0.0009940196247921393, "total_loss": 16.41736602783203, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.93185043334961, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.940971374511719, "koleo_loss": -0.23828125, "ibot_loss": 5.5330986976623535, "backbone_grad_norm": 0.4336176812648773, "dino_head_grad_norm": 0.016334092244505882, "ibot_head_grad_norm": 0.003804302541539073}
|
| 26 |
+
{"iteration": 250, "iter_time": 0.29155975580215454, "data_time": 0.058961689472198486, "lr": 0.0009922210707609008, "wd": 0.059618825646093776, "mom": 0.9924359739032466, "last_layer_lr": 0.0009922210707609008, "total_loss": 16.396709442138672, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.911377906799316, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.922164916992188, "koleo_loss": -0.2392578125, "ibot_loss": 5.53170108795166, "backbone_grad_norm": 0.41960665583610535, "dino_head_grad_norm": 0.014407741837203503, "ibot_head_grad_norm": 0.003942093346267939}
|
| 27 |
+
{"iteration": 260, "iter_time": 0.2419767677783966, "data_time": 0.009889054112136364, "lr": 0.0009901884841315041, "wd": 0.06125417241729608, "mom": 0.9924723149426066, "last_layer_lr": 0.0009901884841315041, "total_loss": 16.378799438476562, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.89584732055664, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.909034729003906, "koleo_loss": -0.2421875, "ibot_loss": 5.529606819152832, "backbone_grad_norm": 0.4403291642665863, "dino_head_grad_norm": 0.015625368803739548, "ibot_head_grad_norm": 0.0040090507827699184}
|
| 28 |
+
{"iteration": 270, "iter_time": 0.2638514041900635, "data_time": 0.032703567296266556, "lr": 0.0009879228323050999, "wd": 0.0629507187268965, "mom": 0.9925100159717088, "last_layer_lr": 0.0009879228323050999, "total_loss": 16.357940673828125, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.876874923706055, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.88817024230957, "koleo_loss": -0.240234375, "ibot_loss": 5.528289794921875, "backbone_grad_norm": 0.4403291642665863, "dino_head_grad_norm": 0.01620626077055931, "ibot_head_grad_norm": 0.00430065905675292}
|
| 29 |
+
{"iteration": 280, "iter_time": 0.3110233247280121, "data_time": 0.07987821102142334, "lr": 0.0009854251936092507, "wd": 0.06470781052399743, "mom": 0.9925490624560889, "last_layer_lr": 0.0009854251936092507, "total_loss": 16.343292236328125, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.863395690917969, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.875669479370117, "koleo_loss": -0.240234375, "ibot_loss": 5.5264387130737305, "backbone_grad_norm": 0.3980690836906433, "dino_head_grad_norm": 0.01872454769909382, "ibot_head_grad_norm": 0.004627150949090719}
|
| 30 |
+
{"iteration": 290, "iter_time": 0.3171479105949402, "data_time": 0.08532911539077759, "lr": 0.0009826967567847036, "wd": 0.06652477041626337, "mom": 0.9925894393425836, "last_layer_lr": 0.0009826967567847036, "total_loss": 16.329072952270508, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.848665237426758, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.86427116394043, "koleo_loss": -0.2392578125, "ibot_loss": 5.525081634521484, "backbone_grad_norm": 0.4212908148765564, "dino_head_grad_norm": 0.020300520583987236, "ibot_head_grad_norm": 0.005068708676844835}
|
| 31 |
+
{"iteration": 300, "iter_time": 0.28770560026168823, "data_time": 0.038768280297517776, "lr": 0.0009797388204196157, "wd": 0.0684008979310684, "mom": 0.9926311310651349, "last_layer_lr": 0.0009797388204196157, "total_loss": 16.319602966308594, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.842247009277344, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.856547355651855, "koleo_loss": -0.2333984375, "ibot_loss": 5.523205757141113, "backbone_grad_norm": 0.514546275138855, "dino_head_grad_norm": 0.02029246650636196, "ibot_head_grad_norm": 0.0052216858603060246}
|
| 32 |
+
{"iteration": 310, "iter_time": 0.2507888972759247, "data_time": 0.0014875292545184493, "lr": 0.0009765527923314955, "wd": 0.07033546978554184, "mom": 0.9926741215507898, "last_layer_lr": 0.0009765527923314955, "total_loss": 16.303237915039062, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.823105812072754, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.831926345825195, "koleo_loss": -0.23046875, "ibot_loss": 5.5208420753479, "backbone_grad_norm": 0.5722232460975647, "dino_head_grad_norm": 0.020487669855356216, "ibot_head_grad_norm": 0.0052216858603060246}
|
| 33 |
+
{"iteration": 320, "iter_time": 0.2331526279449463, "data_time": 0.0009144783252850175, "lr": 0.0009731401888971581, "wd": 0.07232774016540755, "mom": 0.9927183942258979, "last_layer_lr": 0.0009731401888971581, "total_loss": 16.28411865234375, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.807479858398438, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.8257474899292, "koleo_loss": -0.2265625, "ibot_loss": 5.519189834594727, "backbone_grad_norm": 0.5716654062271118, "dino_head_grad_norm": 0.02022848278284073, "ibot_head_grad_norm": 0.005679080728441477}
|
| 34 |
+
{"iteration": 330, "iter_time": 0.25313907861709595, "data_time": 0.021058762446045876, "lr": 0.0009695026343310123, "wd": 0.07437694101250947, "mom": 0.9927639320225002, "last_layer_lr": 0.0009695026343310123, "total_loss": 16.269367218017578, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.795270919799805, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.803462982177734, "koleo_loss": -0.2333984375, "ibot_loss": 5.517642021179199, "backbone_grad_norm": 0.5533215999603271, "dino_head_grad_norm": 0.01797669194638729, "ibot_head_grad_norm": 0.005425119772553444}
|
| 35 |
+
{"iteration": 340, "iter_time": 0.29009732604026794, "data_time": 0.05746457725763321, "lr": 0.0009656418599120225, "wd": 0.0764822823209127, "mom": 0.9928107173849091, "last_layer_lr": 0.0009656418599120225, "total_loss": 16.24939727783203, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.778265953063965, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.791526794433594, "koleo_loss": -0.234375, "ibot_loss": 5.515594482421875, "backbone_grad_norm": 0.652686595916748, "dino_head_grad_norm": 0.01935099996626377, "ibot_head_grad_norm": 0.005324184894561768}
|
| 36 |
+
{"iteration": 350, "iter_time": 0.2778920531272888, "data_time": 0.04575664922595024, "lr": 0.000961559703159713, "wd": 0.07864295244146591, "mom": 0.992858732276477, "last_layer_lr": 0.000961559703159713, "total_loss": 16.23888397216797, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.76366901397705, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.785568237304688, "koleo_loss": -0.228515625, "ibot_loss": 5.513799667358398, "backbone_grad_norm": 0.5551126599311829, "dino_head_grad_norm": 0.023039426654577255, "ibot_head_grad_norm": 0.0060704839415848255}
|
| 37 |
+
{"iteration": 360, "iter_time": 0.24903444945812225, "data_time": 0.017055701464414597, "lr": 0.0009572581069596094, "wd": 0.08085811839470736, "mom": 0.9929079581865491, "last_layer_lr": 0.0009572581069596094, "total_loss": 16.22039031982422, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.752452850341797, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.773161888122559, "koleo_loss": -0.228515625, "ibot_loss": 5.511704444885254, "backbone_grad_norm": 0.5921959280967712, "dino_head_grad_norm": 0.023039426654577255, "ibot_head_grad_norm": 0.0060704839415848255}
|
| 38 |
+
{"iteration": 370, "iter_time": 0.28939586877822876, "data_time": 0.0588475838303566, "lr": 0.0009527391186385288, "wd": 0.08312692619199441, "mom": 0.9929583761375999, "last_layer_lr": 0.0009527391186385288, "total_loss": 16.21061134338379, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.74333381652832, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.755472183227539, "koleo_loss": -0.228515625, "ibot_loss": 5.510573863983154, "backbone_grad_norm": 0.5921959280967712, "dino_head_grad_norm": 0.022539611905813217, "ibot_head_grad_norm": 0.00539066968485713}
|
| 39 |
+
{"iteration": 380, "iter_time": 0.31175750494003296, "data_time": 0.08158449083566666, "lr": 0.0009480048889901648, "wd": 0.08544850116473257, "mom": 0.9930099666925496, "last_layer_lr": 0.0009480048889901648, "total_loss": 16.18854331970215, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.722600936889648, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.738670349121094, "koleo_loss": -0.228515625, "ibot_loss": 5.507286548614502, "backbone_grad_norm": 0.6310003399848938, "dino_head_grad_norm": 0.022576315328478813, "ibot_head_grad_norm": 0.005837167147547007}
|
| 40 |
+
{"iteration": 390, "iter_time": 0.29853057861328125, "data_time": 0.06727175414562225, "lr": 0.000943057671251425, "wd": 0.08782194830157658, "mom": 0.9930627099622572, "last_layer_lr": 0.000943057671251425, "total_loss": 16.1827392578125, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.717269897460938, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.732093811035156, "koleo_loss": -0.2314453125, "ibot_loss": 5.506141185760498, "backbone_grad_norm": 0.7003105878829956, "dino_head_grad_norm": 0.025325296446681023, "ibot_head_grad_norm": 0.0058919214643538}
|
| 41 |
+
{"iteration": 400, "iter_time": 0.2811409533023834, "data_time": 0.04990973323583603, "lr": 0.0009378998200300123, "wd": 0.09024635259347458, "mom": 0.9931165856131883, "last_layer_lr": 0.0009378998200300123, "total_loss": 16.17267608642578, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.711469650268555, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.72297477722168, "koleo_loss": -0.232421875, "ibot_loss": 5.505288124084473, "backbone_grad_norm": 0.6053245663642883, "dino_head_grad_norm": 0.023747486993670464, "ibot_head_grad_norm": 0.005963497329503298}
|
| 42 |
+
{"iteration": 410, "iter_time": 0.24590051174163818, "data_time": 0.014007091522216797, "lr": 0.0009325337901837602, "wd": 0.09272077938642143, "mom": 0.9931715728752538, "last_layer_lr": 0.0009325337901837602, "total_loss": 16.14833641052246, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.687915802001953, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.713916778564453, "koleo_loss": -0.234375, "ibot_loss": 5.503820896148682, "backbone_grad_norm": 0.6309290528297424, "dino_head_grad_norm": 0.02211681194603443, "ibot_head_grad_norm": 0.005736169405281544}
|
| 43 |
+
{"iteration": 420, "iter_time": 0.24009886384010315, "data_time": 0.007970189675688744, "lr": 0.0009269621356522519, "wd": 0.09524427474178554, "mom": 0.9932276505498174, "last_layer_lr": 0.0009269621356522519, "total_loss": 16.12238311767578, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.669026374816895, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.677433013916016, "koleo_loss": -0.2373046875, "ibot_loss": 5.500762939453125, "backbone_grad_norm": 0.5473890900611877, "dino_head_grad_norm": 0.017296135425567627, "ibot_head_grad_norm": 0.005736169405281544}
|
| 44 |
+
{"iteration": 430, "iter_time": 0.2532300055027008, "data_time": 0.021020282059907913, "lr": 0.0009211875082412853, "wd": 0.09781586580407042, "mom": 0.9932847970178682, "last_layer_lr": 0.0009211875082412853, "total_loss": 16.119888305664062, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.662205696105957, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.672443389892578, "koleo_loss": -0.234375, "ibot_loss": 5.499980926513672, "backbone_grad_norm": 0.5398440957069397, "dino_head_grad_norm": 0.02018103189766407, "ibot_head_grad_norm": 0.0070283543318510056}
|
| 45 |
+
{"iteration": 440, "iter_time": 0.2821366786956787, "data_time": 0.04971566051244736, "lr": 0.0009152126563607545, "wd": 0.10043456117596927, "mom": 0.9933429902483548, "last_layer_lr": 0.0009152126563607545, "total_loss": 16.11444854736328, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.65697956085205, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.663816452026367, "koleo_loss": -0.232421875, "ibot_loss": 5.496916770935059, "backbone_grad_norm": 0.564204216003418, "dino_head_grad_norm": 0.022360850125551224, "ibot_head_grad_norm": 0.006263164337724447}
|
| 46 |
+
{"iteration": 450, "iter_time": 0.33341628313064575, "data_time": 0.08480726182460785, "lr": 0.0009090404237165582, "wd": 0.10309935130056691, "mom": 0.9934022078066792, "last_layer_lr": 0.0009090404237165582, "total_loss": 16.095191955566406, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.650904655456543, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.653546333312988, "koleo_loss": -0.234375, "ibot_loss": 5.4962053298950195, "backbone_grad_norm": 0.5107251405715942, "dino_head_grad_norm": 0.0212933998554945, "ibot_head_grad_norm": 0.0065112970769405365}
|
| 47 |
+
{"iteration": 460, "iter_time": 0.31343597173690796, "data_time": 0.06553294509649277, "lr": 0.0009026737479571478, "wd": 0.10580920885054379, "mom": 0.9934624268633454, "last_layer_lr": 0.0009026737479571478, "total_loss": 16.08465576171875, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.630728721618652, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.64529800415039, "koleo_loss": -0.240234375, "ibot_loss": 5.494109630584717, "backbone_grad_norm": 0.5089584589004517, "dino_head_grad_norm": 0.019582506269216537, "ibot_head_grad_norm": 0.007387160789221525}
|
| 48 |
+
{"iteration": 470, "iter_time": 0.2966565787792206, "data_time": 0.06601300090551376, "lr": 0.0008961156592753669, "wd": 0.10856308912422985, "mom": 0.9935236242027606, "last_layer_lr": 0.0008961156592753669, "total_loss": 16.07072639465332, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.620450973510742, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.638181686401367, "koleo_loss": -0.240234375, "ibot_loss": 5.492586612701416, "backbone_grad_norm": 0.5466471314430237, "dino_head_grad_norm": 0.019491400569677353, "ibot_head_grad_norm": 0.0071171727031469345}
|
| 49 |
+
{"iteration": 480, "iter_time": 0.3207061290740967, "data_time": 0.09104973077774048, "lr": 0.0008893692789662431, "wd": 0.11135993044835713, "mom": 0.9935857762321857, "last_layer_lr": 0.0008893692789662431, "total_loss": 16.068025588989258, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.619190216064453, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.629169464111328, "koleo_loss": -0.240234375, "ibot_loss": 5.492668151855469, "backbone_grad_norm": 0.5606192350387573, "dino_head_grad_norm": 0.02056359127163887, "ibot_head_grad_norm": 0.006539354100823402}
|
| 50 |
+
{"iteration": 490, "iter_time": 0.2788059115409851, "data_time": 0.04909191280603409, "lr": 0.0008824378179414209, "wd": 0.11419865458735479, "mom": 0.9936488589908301, "last_layer_lr": 0.0008824378179414209, "total_loss": 16.050682067871094, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.607476234436035, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.613627433776855, "koleo_loss": -0.240234375, "ibot_loss": 5.490269660949707, "backbone_grad_norm": 0.596026599407196, "dino_head_grad_norm": 0.02056359127163887, "ibot_head_grad_norm": 0.007047154475003481}
|
| 51 |
+
{"iteration": 500, "iter_time": 0.23912844061851501, "data_time": 0.007565951440483332, "lr": 0.0008753245752009412, "wd": 0.1170781671590298, "mom": 0.9937128481590896, "last_layer_lr": 0.0008753245752009412, "total_loss": 16.034751892089844, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.596311569213867, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.60368537902832, "koleo_loss": -0.2431640625, "ibot_loss": 5.486662864685059, "backbone_grad_norm": 0.6536314487457275, "dino_head_grad_norm": 0.020832344889640808, "ibot_head_grad_norm": 0.007412055507302284}
|
| 52 |
+
{"iteration": 510, "iter_time": 0.2337232530117035, "data_time": 0.0009210705757141113, "lr": 0.0008680329362630967, "wd": 0.11999735805647155, "mom": 0.9937777190679216, "last_layer_lr": 0.0008680329362630967, "total_loss": 16.027009963989258, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.590981483459473, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.599881172180176, "koleo_loss": -0.2421875, "ibot_loss": 5.484419345855713, "backbone_grad_norm": 0.6150062680244446, "dino_head_grad_norm": 0.020832344889640808, "ibot_head_grad_norm": 0.007038840092718601}
|
| 53 |
+
{"iteration": 520, "iter_time": 0.233208566904068, "data_time": 0.0009226202964782715, "lr": 0.000860566371553108, "wd": 0.12295510187601999, "mom": 0.993843446708356, "last_layer_lr": 0.000860566371553108, "total_loss": 16.011775970458984, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.577241897583008, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.582603454589844, "koleo_loss": -0.23828125, "ibot_loss": 5.480428218841553, "backbone_grad_norm": 0.6017665266990662, "dino_head_grad_norm": 0.021769698709249496, "ibot_head_grad_norm": 0.00714716175571084}
|
| 54 |
+
{"iteration": 530, "iter_time": 0.25649237632751465, "data_time": 0.02476346492767334, "lr": 0.0008529284347513893, "wd": 0.12595025835112922, "mom": 0.9939100057411362, "last_layer_lr": 0.0008529284347513893, "total_loss": 15.99846076965332, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.562795639038086, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.574458122253418, "koleo_loss": -0.2412109375, "ibot_loss": 5.47941255569458, "backbone_grad_norm": 0.6178594827651978, "dino_head_grad_norm": 0.022689897567033768, "ibot_head_grad_norm": 0.008532549254596233}
|
| 55 |
+
{"iteration": 540, "iter_time": 0.29501408338546753, "data_time": 0.06384921073913574, "lr": 0.0008451227611021889, "wd": 0.12898167279196276, "mom": 0.993977370506488, "last_layer_lr": 0.0008451227611021889, "total_loss": 15.985437393188477, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.558700561523438, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.560848236083984, "koleo_loss": -0.2412109375, "ibot_loss": 5.475007057189941, "backbone_grad_norm": 0.596586287021637, "dino_head_grad_norm": 0.021747998893260956, "ibot_head_grad_norm": 0.008493577130138874}
|
| 56 |
+
{"iteration": 550, "iter_time": 0.31851521134376526, "data_time": 0.08826988935470581, "lr": 0.00083715306568341, "wd": 0.13204817653054807, "mom": 0.9940455150340122, "last_layer_lr": 0.00083715306568341, "total_loss": 15.968608856201172, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.541694641113281, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.542327880859375, "koleo_loss": -0.244140625, "ibot_loss": 5.471616744995117, "backbone_grad_norm": 0.6213814616203308, "dino_head_grad_norm": 0.019885655492544174, "ibot_head_grad_norm": 0.007795560173690319}
|
| 57 |
+
{"iteration": 560, "iter_time": 0.3158690929412842, "data_time": 0.08547288179397583, "lr": 0.0008290231416384344, "wd": 0.13514858737132046, "mom": 0.994114413052696, "last_layer_lr": 0.0008290231416384344, "total_loss": 15.952314376831055, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.52934741973877, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.542786598205566, "koleo_loss": -0.2451171875, "ibot_loss": 5.469967842102051, "backbone_grad_norm": 0.5412294268608093, "dino_head_grad_norm": 0.019885655492544174, "ibot_head_grad_norm": 0.00891039613634348}
|
| 58 |
+
{"iteration": 570, "iter_time": 0.28711000084877014, "data_time": 0.05575763061642647, "lr": 0.0008207368583707927, "wd": 0.13828171004688156, "mom": 0.9941840380010418, "last_layer_lr": 0.0008207368583707927, "total_loss": 15.947593688964844, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.528284072875977, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.546215057373047, "koleo_loss": -0.244140625, "ibot_loss": 5.468135833740234, "backbone_grad_norm": 0.7932685613632202, "dino_head_grad_norm": 0.02151632122695446, "ibot_head_grad_norm": 0.009271892718970776}
|
| 59 |
+
{"iteration": 580, "iter_time": 0.2805480360984802, "data_time": 0.04926940053701401, "lr": 0.0008122981597025365, "wd": 0.1414463366787984, "mom": 0.9942543630373066, "last_layer_lr": 0.0008122981597025365, "total_loss": 15.926070213317871, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.506810188293457, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.516460418701172, "koleo_loss": -0.23828125, "ibot_loss": 5.463968276977539, "backbone_grad_norm": 0.800346314907074, "dino_head_grad_norm": 0.02222329005599022, "ibot_head_grad_norm": 0.008340511471033096}
|
| 60 |
+
{"iteration": 590, "iter_time": 0.3093538284301758, "data_time": 0.07921556383371353, "lr": 0.0008037110619971935, "wd": 0.14464124724326294, "mom": 0.9943253610498503, "last_layer_lr": 0.0008037110619971935, "total_loss": 15.914482116699219, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.501279830932617, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.515562057495117, "koleo_loss": -0.23828125, "ibot_loss": 5.460835933685303, "backbone_grad_norm": 0.6548895835876465, "dino_head_grad_norm": 0.022131511941552162, "ibot_head_grad_norm": 0.00778606254607439}
|
| 61 |
+
{"iteration": 600, "iter_time": 0.3081631064414978, "data_time": 0.06138354539871216, "lr": 0.0007949796522481959, "wd": 0.14786521004143466, "mom": 0.9943970046675874, "last_layer_lr": 0.0007949796522481959, "total_loss": 15.90496826171875, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.49679946899414, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.50379753112793, "koleo_loss": -0.2421875, "ibot_loss": 5.456981182098389, "backbone_grad_norm": 0.6548895835876465, "dino_head_grad_norm": 0.021260233595967293, "ibot_head_grad_norm": 0.008621146902441978}
|
| 62 |
+
{"iteration": 610, "iter_time": 0.25961780548095703, "data_time": 0.012831282801926136, "lr": 0.0007861080861336924, "wd": 0.15111698217428382, "mom": 0.9944692662705397, "last_layer_lr": 0.0007861080861336924, "total_loss": 15.882318496704102, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.47879409790039, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.489937782287598, "koleo_loss": -0.244140625, "ibot_loss": 5.453865051269531, "backbone_grad_norm": 0.6043180227279663, "dino_head_grad_norm": 0.022143159061670303, "ibot_head_grad_norm": 0.008799968287348747}
|
| 63 |
+
{"iteration": 620, "iter_time": 0.22791704535484314, "data_time": 0.0007244705921038985, "lr": 0.0007771005860386714, "wd": 0.15439531002175308, "mom": 0.9945421180004834, "last_layer_lr": 0.0007771005860386714, "total_loss": 15.866752624511719, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.460265159606934, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.47551441192627, "koleo_loss": -0.244140625, "ibot_loss": 5.452833652496338, "backbone_grad_norm": 0.6043180227279663, "dino_head_grad_norm": 0.02319164015352726, "ibot_head_grad_norm": 0.009720607660710812}
|
| 64 |
+
{"iteration": 630, "iter_time": 0.2255133092403412, "data_time": 0.0006253600004129112, "lr": 0.0007679614390453336, "wd": 0.15769892972605126, "mom": 0.99461553177169, "last_layer_lr": 0.0007679614390453336, "total_loss": 15.857216835021973, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.45352840423584, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.460857391357422, "koleo_loss": -0.24609375, "ibot_loss": 5.451993942260742, "backbone_grad_norm": 0.55731201171875, "dino_head_grad_norm": 0.021913569420576096, "ibot_head_grad_norm": 0.010125784203410149}
|
| 65 |
+
{"iteration": 640, "iter_time": 0.23831510543823242, "data_time": 0.013437176123261452, "lr": 0.0007586949948926728, "wd": 0.1610265676788952, "mom": 0.9946894792817532, "last_layer_lr": 0.0007586949948926728, "total_loss": 15.84133529663086, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.446563720703125, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.454069137573242, "koleo_loss": -0.25, "ibot_loss": 5.4458909034729, "backbone_grad_norm": 0.6424360275268555, "dino_head_grad_norm": 0.020732201635837555, "ibot_head_grad_norm": 0.008863376453518867}
|
| 66 |
+
{"iteration": 650, "iter_time": 0.2948307394981384, "data_time": 0.06744764745235443, "lr": 0.0007493056639062373, "wd": 0.16437694101250946, "mom": 0.9947639320225002, "last_layer_lr": 0.0007493056639062373, "total_loss": 15.827220916748047, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.428558349609375, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.433884620666504, "koleo_loss": -0.25, "ibot_loss": 5.445651054382324, "backbone_grad_norm": 0.6551433801651001, "dino_head_grad_norm": 0.020732201635837555, "ibot_head_grad_norm": 0.009534113109111786}
|
| 67 |
+
{"iteration": 660, "iter_time": 0.31735119223594666, "data_time": 0.08662009239196777, "lr": 0.000739797914899052, "wd": 0.16774875809419673, "mom": 0.9948388612909821, "last_layer_lr": 0.000739797914899052, "total_loss": 15.816849708557129, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.422212600708008, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.427522659301758, "koleo_loss": -0.244140625, "ibot_loss": 5.440255165100098, "backbone_grad_norm": 0.6732689738273621, "dino_head_grad_norm": 0.024994971230626106, "ibot_head_grad_norm": 0.011124961078166962}
|
| 68 |
+
{"iteration": 670, "iter_time": 0.27196693420410156, "data_time": 0.04023389890789986, "lr": 0.0007301762730447061, "wd": 0.17114071902428657, "mom": 0.9949142382005397, "last_layer_lr": 0.0007301762730447061, "total_loss": 15.801877975463867, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.415157318115234, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.417402267456055, "koleo_loss": -0.248046875, "ibot_loss": 5.43771505355835, "backbone_grad_norm": 0.6752502918243408, "dino_head_grad_norm": 0.024733366444706917, "ibot_head_grad_norm": 0.009969576261937618}
|
| 69 |
+
{"iteration": 680, "iter_time": 0.24158553779125214, "data_time": 0.00902494229376316, "lr": 0.0007204453177236162, "wd": 0.17455151613727155, "mom": 0.9949900336919394, "last_layer_lr": 0.0007204453177236162, "total_loss": 15.79538345336914, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.408700942993164, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.417402267456055, "koleo_loss": -0.2490234375, "ibot_loss": 5.434782981872559, "backbone_grad_norm": 0.6022997498512268, "dino_head_grad_norm": 0.022493712604045868, "ibot_head_grad_norm": 0.010746297426521778}
|
| 70 |
+
{"iteration": 690, "iter_time": 0.24943479895591736, "data_time": 0.016623079776763916, "lr": 0.0007106096803434884, "wd": 0.177979834505937, "mom": 0.9950662185445763, "last_layer_lr": 0.0007106096803434884, "total_loss": 15.770469665527344, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.387154579162598, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.408014297485352, "koleo_loss": -0.25, "ibot_loss": 5.4295501708984375, "backbone_grad_norm": 0.5990486145019531, "dino_head_grad_norm": 0.020357534289360046, "ibot_head_grad_norm": 0.011523088440299034}
|
| 71 |
+
{"iteration": 700, "iter_time": 0.29624396562576294, "data_time": 0.06338337808847427, "lr": 0.0007006740421350191, "wd": 0.18142435244829086, "mom": 0.9951427633877398, "last_layer_lr": 0.0007006740421350191, "total_loss": 15.760028839111328, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.378271102905273, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.392601013183594, "koleo_loss": -0.251953125, "ibot_loss": 5.427119255065918, "backbone_grad_norm": 0.6032534837722778, "dino_head_grad_norm": 0.024071337655186653, "ibot_head_grad_norm": 0.011665534228086472}
|
| 72 |
+
{"iteration": 710, "iter_time": 0.3272583484649658, "data_time": 0.09495820105075836, "lr": 0.0006906431319238832, "wd": 0.18488374203709687, "mom": 0.9952196387119355, "last_layer_lr": 0.0006906431319238832, "total_loss": 15.752090454101562, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.376689910888672, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.377927780151367, "koleo_loss": -0.25, "ibot_loss": 5.423197269439697, "backbone_grad_norm": 0.7946348786354065, "dino_head_grad_norm": 0.02210947871208191, "ibot_head_grad_norm": 0.012578101828694344}
|
| 73 |
+
{"iteration": 720, "iter_time": 0.2903120517730713, "data_time": 0.06033952161669731, "lr": 0.0006805217238800692, "wd": 0.18835666961181616, "mom": 0.9952968148802626, "last_layer_lr": 0.0006805217238800692, "total_loss": 15.742185592651367, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.37452220916748, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.371755599975586, "koleo_loss": -0.25, "ibot_loss": 5.420013427734375, "backbone_grad_norm": 0.5722468495368958, "dino_head_grad_norm": 0.020922740921378136, "ibot_head_grad_norm": 0.012588437646627426}
|
| 74 |
+
{"iteration": 730, "iter_time": 0.2441457211971283, "data_time": 0.013697469606995583, "lr": 0.0006703146352456317, "wd": 0.19184179629275844, "mom": 0.995374262139839, "last_layer_lr": 0.0006703146352456317, "total_loss": 15.724117279052734, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.360414505004883, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.363025665283203, "koleo_loss": -0.251953125, "ibot_loss": 5.41574764251709, "backbone_grad_norm": 0.5722468495368958, "dino_head_grad_norm": 0.022787820547819138, "ibot_head_grad_norm": 0.01187440101057291}
|
| 75 |
+
{"iteration": 740, "iter_time": 0.23452934622764587, "data_time": 0.0023980974219739437, "lr": 0.0006600267240419474, "wd": 0.19533777849724576, "mom": 0.9954519506332721, "last_layer_lr": 0.0006600267240419474, "total_loss": 15.717137336730957, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.347832679748535, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.360288619995117, "koleo_loss": -0.25, "ibot_loss": 5.411498069763184, "backbone_grad_norm": 0.6575363874435425, "dino_head_grad_norm": 0.025500338524580002, "ibot_head_grad_norm": 0.0110955024138093}
|
| 76 |
+
{"iteration": 750, "iter_time": 0.2653677463531494, "data_time": 0.017271708697080612, "lr": 0.0006496628867575574, "wd": 0.1988432684575892, "mom": 0.9955298504101686, "last_layer_lr": 0.0006496628867575574, "total_loss": 15.705204010009766, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.34774398803711, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.359582901000977, "koleo_loss": -0.24609375, "ibot_loss": 5.408249855041504, "backbone_grad_norm": 0.5417140126228333, "dino_head_grad_norm": 0.025083474814891815, "ibot_head_grad_norm": 0.010904286988079548}
|
| 77 |
+
{"iteration": 760, "iter_time": 0.302177369594574, "data_time": 0.055673517286777496, "lr": 0.0006392280560177056, "wd": 0.20235691474067904, "mom": 0.9956079314386818, "last_layer_lr": 0.0006392280560177056, "total_loss": 15.691631317138672, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.339563369750977, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.351705551147461, "koleo_loss": -0.251953125, "ibot_loss": 5.404703617095947, "backbone_grad_norm": 0.4952976703643799, "dino_head_grad_norm": 0.022777538746595383, "ibot_head_grad_norm": 0.011373286135494709}
|
| 78 |
+
{"iteration": 770, "iter_time": 0.3083978295326233, "data_time": 0.08076868951320648, "lr": 0.0006287271982366756, "wd": 0.20587736276898788, "mom": 0.9956861636170886, "last_layer_lr": 0.0006287271982366756, "total_loss": 15.670628547668457, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.323238372802734, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.325642585754395, "koleo_loss": -0.25390625, "ibot_loss": 5.401246547698975, "backbone_grad_norm": 0.5548362135887146, "dino_head_grad_norm": 0.02080795355141163, "ibot_head_grad_norm": 0.011373286135494709}
|
| 79 |
+
{"iteration": 780, "iter_time": 0.29322919249534607, "data_time": 0.0631730705499649, "lr": 0.0006181653112540469, "wd": 0.209403255342786, "mom": 0.9957645167853952, "last_layer_lr": 0.0006181653112540469, "total_loss": 15.658540725708008, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.311100959777832, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.314653396606445, "koleo_loss": -0.25390625, "ibot_loss": 5.3965229988098145, "backbone_grad_norm": 0.5707761645317078, "dino_head_grad_norm": 0.021036317571997643, "ibot_head_grad_norm": 0.01209683995693922}
|
| 80 |
+
{"iteration": 790, "iter_time": 0.2754022777080536, "data_time": 0.04225137084722519, "lr": 0.0006075474219559929, "wd": 0.21293323316336765, "mom": 0.9958429607369638, "last_layer_lr": 0.0006075474219559929, "total_loss": 15.643972396850586, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.304252624511719, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.311185836791992, "koleo_loss": -0.2578125, "ibot_loss": 5.392583847045898, "backbone_grad_norm": 0.5261061191558838, "dino_head_grad_norm": 0.023725291714072227, "ibot_head_grad_norm": 0.012943794950842857}
|
| 81 |
+
{"iteration": 800, "iter_time": 0.25236159563064575, "data_time": 0.019969046115875244, "lr": 0.0005968785838827569, "wd": 0.21646593535708689, "mom": 0.9959214652301575, "last_layer_lr": 0.0005968785838827569, "total_loss": 15.639662742614746, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.303186416625977, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.309297561645508, "koleo_loss": -0.25390625, "ibot_loss": 5.389980792999268, "backbone_grad_norm": 0.545911431312561, "dino_head_grad_norm": 0.023725291714072227, "ibot_head_grad_norm": 0.011734903790056705}
|
| 82 |
+
{"iteration": 810, "iter_time": 0.23881463706493378, "data_time": 0.006575083825737238, "lr": 0.0005861638748234393, "wd": 0.22, "mom": 0.996, "last_layer_lr": 0.0005861638748234393, "total_loss": 15.631072998046875, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.296869277954102, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.295449256896973, "koleo_loss": -0.251953125, "ibot_loss": 5.388440132141113, "backbone_grad_norm": 0.6017118692398071, "dino_head_grad_norm": 0.024190833792090416, "ibot_head_grad_norm": 0.014039311558008194}
|
| 83 |
+
{"iteration": 820, "iter_time": 0.26122236251831055, "data_time": 0.029075611382722855, "lr": 0.0005754083943992455, "wd": 0.22353406464291306, "mom": 0.9960785347698425, "last_layer_lr": 0.0005754083943992455, "total_loss": 15.613727569580078, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.28119945526123, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.281405448913574, "koleo_loss": -0.25390625, "ibot_loss": 5.382573127746582, "backbone_grad_norm": 0.5546561479568481, "dino_head_grad_norm": 0.023393932729959488, "ibot_head_grad_norm": 0.014640973880887032}
|
| 84 |
+
{"iteration": 830, "iter_time": 0.2772912383079529, "data_time": 0.04513983801007271, "lr": 0.0005646172616363428, "wd": 0.22706676683663235, "mom": 0.9961570392630362, "last_layer_lr": 0.0005646172616363428, "total_loss": 15.612147331237793, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.280237197875977, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.281301498413086, "koleo_loss": -0.255859375, "ibot_loss": 5.37962007522583, "backbone_grad_norm": 0.49813172221183777, "dino_head_grad_norm": 0.022516200318932533, "ibot_head_grad_norm": 0.01248974446207285}
|
| 85 |
+
{"iteration": 840, "iter_time": 0.27506357431411743, "data_time": 0.043200064450502396, "lr": 0.0005537956125294805, "wd": 0.23059674465721405, "mom": 0.9962354832146048, "last_layer_lr": 0.0005537956125294805, "total_loss": 15.60373306274414, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.276750564575195, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.279325485229492, "koleo_loss": -0.2578125, "ibot_loss": 5.37868070602417, "backbone_grad_norm": 0.5698763132095337, "dino_head_grad_norm": 0.02183677814900875, "ibot_head_grad_norm": 0.012365789152681828}
|
| 86 |
+
{"iteration": 850, "iter_time": 0.27372026443481445, "data_time": 0.042191945016384125, "lr": 0.0005429485975975347, "wd": 0.23412263723101207, "mom": 0.9963138363829114, "last_layer_lr": 0.0005429485975975347, "total_loss": 15.58674430847168, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.263544082641602, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.270763397216797, "koleo_loss": -0.2578125, "ibot_loss": 5.375421524047852, "backbone_grad_norm": 0.5805675387382507, "dino_head_grad_norm": 0.02183677814900875, "ibot_head_grad_norm": 0.012446138076484203}
|
| 87 |
+
{"iteration": 860, "iter_time": 0.26990240812301636, "data_time": 0.0381741039454937, "lr": 0.0005320813794321409, "wd": 0.23764308525932087, "mom": 0.9963920685613182, "last_layer_lr": 0.0005320813794321409, "total_loss": 15.570982933044434, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.254465103149414, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.259326934814453, "koleo_loss": -0.2578125, "ibot_loss": 5.373359203338623, "backbone_grad_norm": 0.5437816977500916, "dino_head_grad_norm": 0.02150239609181881, "ibot_head_grad_norm": 0.013054249808192253}
|
| 88 |
+
{"iteration": 870, "iter_time": 0.27108263969421387, "data_time": 0.03919566795229912, "lr": 0.0005211991302405791, "wd": 0.24115673154241077, "mom": 0.9964701495898314, "last_layer_lr": 0.0005211991302405791, "total_loss": 15.567240715026855, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.249068260192871, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.244878768920898, "koleo_loss": -0.2578125, "ibot_loss": 5.367486953735352, "backbone_grad_norm": 0.5437816977500916, "dino_head_grad_norm": 0.02087969146668911, "ibot_head_grad_norm": 0.013316931203007698}
|
| 89 |
+
{"iteration": 880, "iter_time": 0.2742127776145935, "data_time": 0.04176940768957138, "lr": 0.0005103070293840839, "wd": 0.24466222150275427, "mom": 0.9965480493667279, "last_layer_lr": 0.0005103070293840839, "total_loss": 15.563702583312988, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.248603820800781, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.245824813842773, "koleo_loss": -0.2578125, "ibot_loss": 5.366499900817871, "backbone_grad_norm": 0.5655385851860046, "dino_head_grad_norm": 0.02189713902771473, "ibot_head_grad_norm": 0.013505535200238228}
|
| 90 |
+
{"iteration": 890, "iter_time": 0.26463937759399414, "data_time": 0.03226292133331299, "lr": 0.0004994102609127489, "wd": 0.24815820370724156, "mom": 0.996625737860161, "last_layer_lr": 0.0004994102609127489, "total_loss": 15.546096801757812, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.240872383117676, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.245086669921875, "koleo_loss": -0.263671875, "ibot_loss": 5.360263824462891, "backbone_grad_norm": 0.5655385851860046, "dino_head_grad_norm": 0.02189713902771473, "ibot_head_grad_norm": 0.012421865016222}
|
| 91 |
+
{"iteration": 900, "iter_time": 0.2830435633659363, "data_time": 0.0346997007727623, "lr": 0.0004885140110981993, "wd": 0.25164333038818376, "mom": 0.9967031851197374, "last_layer_lr": 0.0004885140110981993, "total_loss": 15.535469055175781, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.226566314697266, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.233386993408203, "koleo_loss": -0.263671875, "ibot_loss": 5.357806205749512, "backbone_grad_norm": 0.5328091382980347, "dino_head_grad_norm": 0.02066001668572426, "ibot_head_grad_norm": 0.012421865016222}
|
| 92 |
+
{"iteration": 910, "iter_time": 0.3024410605430603, "data_time": 0.05356564372777939, "lr": 0.0004776234659652079, "wd": 0.25511625796290305, "mom": 0.9967803612880645, "last_layer_lr": 0.0004776234659652079, "total_loss": 15.5277738571167, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.224281311035156, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.231666564941406, "koleo_loss": -0.26171875, "ibot_loss": 5.358263969421387, "backbone_grad_norm": 0.568397045135498, "dino_head_grad_norm": 0.022011272609233856, "ibot_head_grad_norm": 0.014166083186864853}
|
| 93 |
+
{"iteration": 920, "iter_time": 0.3090893626213074, "data_time": 0.07695093005895615, "lr": 0.0004667438088234284, "wd": 0.25857564755170914, "mom": 0.9968572366122602, "last_layer_lr": 0.0004667438088234284, "total_loss": 15.516207695007324, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.211904525756836, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.21904182434082, "koleo_loss": -0.265625, "ibot_loss": 5.355380058288574, "backbone_grad_norm": 0.6017688512802124, "dino_head_grad_norm": 0.02303597703576088, "ibot_head_grad_norm": 0.014166083186864853}
|
| 94 |
+
{"iteration": 930, "iter_time": 0.29681578278541565, "data_time": 0.06510509550571442, "lr": 0.0004558802178004195, "wd": 0.26202016549406304, "mom": 0.9969337814554237, "last_layer_lr": 0.0004558802178004195, "total_loss": 15.506583213806152, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.211904525756836, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.207859992980957, "koleo_loss": -0.26171875, "ibot_loss": 5.347967147827148, "backbone_grad_norm": 0.7098405957221985, "dino_head_grad_norm": 0.02357291430234909, "ibot_head_grad_norm": 0.012635529972612858}
|
| 95 |
+
{"iteration": 940, "iter_time": 0.30577975511550903, "data_time": 0.07634687423706055, "lr": 0.0004450378633771384, "wd": 0.2654484838627284, "mom": 0.9970099663080606, "last_layer_lr": 0.0004450378633771384, "total_loss": 15.505399703979492, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.201926231384277, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.197982788085938, "koleo_loss": -0.26171875, "ibot_loss": 5.342251777648926, "backbone_grad_norm": 0.60326087474823, "dino_head_grad_norm": 0.02317691780626774, "ibot_head_grad_norm": 0.014220179989933968}
|
| 96 |
+
{"iteration": 950, "iter_time": 0.31053024530410767, "data_time": 0.08225999027490616, "lr": 0.00043422190592707044, "wd": 0.2688592809757134, "mom": 0.9970857617994603, "last_layer_lr": 0.00043422190592707044, "total_loss": 15.500045776367188, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.195761680603027, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.192826271057129, "koleo_loss": -0.263671875, "ibot_loss": 5.342523574829102, "backbone_grad_norm": 0.48057398200035095, "dino_head_grad_norm": 0.02380499057471752, "ibot_head_grad_norm": 0.016250919550657272}
|
| 97 |
+
{"iteration": 960, "iter_time": 0.2942430078983307, "data_time": 0.06361202150583267, "lr": 0.00042343749326017194, "wd": 0.27225124190580324, "mom": 0.9971611387090178, "last_layer_lr": 0.00042343749326017194, "total_loss": 15.493965148925781, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.194829940795898, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.198639869689941, "koleo_loss": -0.263671875, "ibot_loss": 5.342250823974609, "backbone_grad_norm": 0.48057398200035095, "dino_head_grad_norm": 0.022863641381263733, "ibot_head_grad_norm": 0.016141071915626526}
|
| 98 |
+
{"iteration": 970, "iter_time": 0.2996998727321625, "data_time": 0.07102151960134506, "lr": 0.0004126897581727901, "wd": 0.2756230589874905, "mom": 0.9972360679774998, "last_layer_lr": 0.0004126897581727901, "total_loss": 15.475908279418945, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.193380355834961, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.198639869689941, "koleo_loss": -0.265625, "ibot_loss": 5.338992595672607, "backbone_grad_norm": 0.502232015132904, "dino_head_grad_norm": 0.02254525013267994, "ibot_head_grad_norm": 0.016141071915626526}
|
| 99 |
+
{"iteration": 980, "iter_time": 0.29628506302833557, "data_time": 0.0700007900595665, "lr": 0.00040198381600473114, "wd": 0.27897343232110483, "mom": 0.9973105207182468, "last_layer_lr": 0.00040198381600473114, "total_loss": 15.470867156982422, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.185121536254883, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.192138671875, "koleo_loss": -0.265625, "ibot_loss": 5.33814001083374, "backbone_grad_norm": 0.5159247517585754, "dino_head_grad_norm": 0.02254525013267994, "ibot_head_grad_norm": 0.01670653373003006}
|
| 100 |
+
{"iteration": 990, "iter_time": 0.286446213722229, "data_time": 0.05696103721857071, "lr": 0.0003913247622046356, "wd": 0.28230107027394874, "mom": 0.99738446822831, "last_layer_lr": 0.0003913247622046356, "total_loss": 15.460341453552246, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.179792404174805, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.178979873657227, "koleo_loss": -0.265625, "ibot_loss": 5.335657119750977, "backbone_grad_norm": 0.4861811697483063, "dino_head_grad_norm": 0.020451083779335022, "ibot_head_grad_norm": 0.01629721000790596}
|
| 101 |
+
{"iteration": 1000, "iter_time": 0.2596895098686218, "data_time": 0.02741607464849949, "lr": 0.0003807176699048193, "wd": 0.28560468997824695, "mom": 0.9974578819995166, "last_layer_lr": 0.0003807176699048193, "total_loss": 15.458398818969727, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.175846099853516, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.178979873657227, "koleo_loss": -0.265625, "ibot_loss": 5.333711624145508, "backbone_grad_norm": 0.5053253769874573, "dino_head_grad_norm": 0.02034975402057171, "ibot_head_grad_norm": 0.015619185753166676}
|
| 102 |
+
{"iteration": 1010, "iter_time": 0.24001507461071014, "data_time": 0.008023845963180065, "lr": 0.00037016758750673924, "wd": 0.2888830178257161, "mom": 0.9975307337294603, "last_layer_lr": 0.00037016758750673924, "total_loss": 15.44647216796875, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.166759490966797, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.176464080810547, "koleo_loss": -0.265625, "ibot_loss": 5.329834461212158, "backbone_grad_norm": 0.5256214141845703, "dino_head_grad_norm": 0.02154611051082611, "ibot_head_grad_norm": 0.01610206440091133}
|
| 103 |
+
{"iteration": 1020, "iter_time": 0.23380127549171448, "data_time": 0.0016073227161541581, "lr": 0.0003596795362782263, "wd": 0.29213478995856534, "mom": 0.9976029953324126, "last_layer_lr": 0.0003596795362782263, "total_loss": 15.433408737182617, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.162266731262207, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.156402587890625, "koleo_loss": -0.265625, "ibot_loss": 5.329834461212158, "backbone_grad_norm": 0.4936876893043518, "dino_head_grad_norm": 0.02320987731218338, "ibot_head_grad_norm": 0.0168041680008173}
|
| 104 |
+
{"iteration": 1030, "iter_time": 0.25031575560569763, "data_time": 0.017479311674833298, "lr": 0.0003492585079636341, "wd": 0.29535875275673706, "mom": 0.9976746389501497, "last_layer_lr": 0.0003492585079636341, "total_loss": 15.43262767791748, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.159366607666016, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.155315399169922, "koleo_loss": -0.265625, "ibot_loss": 5.322314262390137, "backbone_grad_norm": 0.4936876893043518, "dino_head_grad_norm": 0.019947806373238564, "ibot_head_grad_norm": 0.0168041680008173}
|
| 105 |
+
{"iteration": 1040, "iter_time": 0.2854323387145996, "data_time": 0.053277600556612015, "lr": 0.00033890946240803957, "wd": 0.2985536633212016, "mom": 0.9977456369626934, "last_layer_lr": 0.00033890946240803957, "total_loss": 15.430773735046387, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.163993835449219, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.170209884643555, "koleo_loss": -0.265625, "ibot_loss": 5.3185882568359375, "backbone_grad_norm": 0.4806305468082428, "dino_head_grad_norm": 0.01940924860537052, "ibot_head_grad_norm": 0.013881168328225613}
|
| 106 |
+
{"iteration": 1050, "iter_time": 0.28963425755500793, "data_time": 0.04137233644723892, "lr": 0.00032863732519662446, "wd": 0.30171828995311845, "mom": 0.9978159619989582, "last_layer_lr": 0.00032863732519662446, "total_loss": 15.422920227050781, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.155839920043945, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.154619216918945, "koleo_loss": -0.26953125, "ibot_loss": 5.3179755210876465, "backbone_grad_norm": 0.45066532492637634, "dino_head_grad_norm": 0.021462617442011833, "ibot_head_grad_norm": 0.014742640778422356}
|
| 107 |
+
{"iteration": 1060, "iter_time": 0.2787701189517975, "data_time": 0.03293421119451523, "lr": 0.00031844698531036477, "wd": 0.3048514126286796, "mom": 0.997885586947304, "last_layer_lr": 0.00031844698531036477, "total_loss": 15.409866333007812, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.146751403808594, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.148868560791016, "koleo_loss": -0.26953125, "ibot_loss": 5.316269397735596, "backbone_grad_norm": 0.45066532492637634, "dino_head_grad_norm": 0.020967848598957062, "ibot_head_grad_norm": 0.015753166750073433}
|
| 108 |
+
{"iteration": 1070, "iter_time": 0.3055942952632904, "data_time": 0.0760219544172287, "lr": 0.0003083432927991392, "wd": 0.30795182346945194, "mom": 0.9979544849659878, "last_layer_lr": 0.0003083432927991392, "total_loss": 15.401796340942383, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.145035743713379, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.146928787231445, "koleo_loss": -0.26953125, "ibot_loss": 5.3108720779418945, "backbone_grad_norm": 0.46550774574279785, "dino_head_grad_norm": 0.020688919350504875, "ibot_head_grad_norm": 0.015120087191462517}
|
| 109 |
+
{"iteration": 1080, "iter_time": 0.29132789373397827, "data_time": 0.05908653885126114, "lr": 0.0002983310564733688, "wd": 0.31101832720803724, "mom": 0.998022629493512, "last_layer_lr": 0.0002983310564733688, "total_loss": 15.39900016784668, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.145035743713379, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.139705657958984, "koleo_loss": -0.26953125, "ibot_loss": 5.310504913330078, "backbone_grad_norm": 0.5278182625770569, "dino_head_grad_norm": 0.02027311734855175, "ibot_head_grad_norm": 0.016068726778030396}
|
| 110 |
+
{"iteration": 1090, "iter_time": 0.26076656579971313, "data_time": 0.028587568551301956, "lr": 0.0002884150416152838, "wd": 0.3140497416488708, "mom": 0.9980899942588638, "last_layer_lr": 0.0002884150416152838, "total_loss": 15.391441345214844, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.139169692993164, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.135828018188477, "koleo_loss": -0.26953125, "ibot_loss": 5.309739589691162, "backbone_grad_norm": 0.5390815734863281, "dino_head_grad_norm": 0.02027311734855175, "ibot_head_grad_norm": 0.014563607051968575}
|
| 111 |
+
{"iteration": 1100, "iter_time": 0.2551064193248749, "data_time": 0.022605489939451218, "lr": 0.0002785999677109072, "wd": 0.31704489812398007, "mom": 0.998156553291644, "last_layer_lr": 0.0002785999677109072, "total_loss": 15.388309478759766, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.128389358520508, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.132904052734375, "koleo_loss": -0.267578125, "ibot_loss": 5.310454368591309, "backbone_grad_norm": 0.5099591612815857, "dino_head_grad_norm": 0.019050322473049164, "ibot_head_grad_norm": 0.013460149988532066}
|
| 112 |
+
{"iteration": 1110, "iter_time": 0.2644177973270416, "data_time": 0.03161132335662842, "lr": 0.000268890506203835, "wd": 0.32000264194352845, "mom": 0.9982222809320784, "last_layer_lr": 0.000268890506203835, "total_loss": 15.38670539855957, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.129725456237793, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.131163597106934, "koleo_loss": -0.26953125, "ibot_loss": 5.308288097381592, "backbone_grad_norm": 0.5099591612815857, "dino_head_grad_norm": 0.020074311643838882, "ibot_head_grad_norm": 0.015160678885877132}
|
| 113 |
+
{"iteration": 1120, "iter_time": 0.3080586791038513, "data_time": 0.07504161447286606, "lr": 0.0002592912782718814, "wd": 0.32292183284097026, "mom": 0.9982871518409104, "last_layer_lr": 0.0002592912782718814, "total_loss": 15.379914283752441, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.131755828857422, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.12624740600586, "koleo_loss": -0.26953125, "ibot_loss": 5.308288097381592, "backbone_grad_norm": 0.4585115909576416, "dino_head_grad_norm": 0.020333902910351753, "ibot_head_grad_norm": 0.014916280284523964}
|
| 114 |
+
{"iteration": 1130, "iter_time": 0.3279266357421875, "data_time": 0.09762952476739883, "lr": 0.00024980685262764995, "wd": 0.32580134541264516, "mom": 0.9983511410091699, "last_layer_lr": 0.00024980685262764995, "total_loss": 15.372261047363281, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.12220573425293, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.123066902160645, "koleo_loss": -0.26953125, "ibot_loss": 5.3050432205200195, "backbone_grad_norm": 0.4585115909576416, "dino_head_grad_norm": 0.020380252972245216, "ibot_head_grad_norm": 0.013666474260389805}
|
| 115 |
+
{"iteration": 1140, "iter_time": 0.27953189611434937, "data_time": 0.05197175592184067, "lr": 0.00024044174334407248, "wd": 0.3286400695516429, "mom": 0.9984142237678143, "last_layer_lr": 0.00024044174334407248, "total_loss": 15.362272262573242, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.115955352783203, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.119512557983398, "koleo_loss": -0.271484375, "ibot_loss": 5.298760414123535, "backbone_grad_norm": 0.40716931223869324, "dino_head_grad_norm": 0.019654501229524612, "ibot_head_grad_norm": 0.013090669177472591}
|
| 116 |
+
{"iteration": 1150, "iter_time": 0.2532022297382355, "data_time": 0.026315461844205856, "lr": 0.00023120040770595538, "wd": 0.33143691087577015, "mom": 0.9984763757972394, "last_layer_lr": 0.00023120040770595538, "total_loss": 15.364984512329102, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.115955352783203, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.11917781829834, "koleo_loss": -0.2734375, "ibot_loss": 5.300388336181641, "backbone_grad_norm": 0.41288816928863525, "dino_head_grad_norm": 0.019910108298063278, "ibot_head_grad_norm": 0.014221868477761745}
|
| 117 |
+
{"iteration": 1160, "iter_time": 0.3006346523761749, "data_time": 0.07344422489404678, "lr": 0.00022208724408855331, "wd": 0.33419079114945616, "mom": 0.9985375731366546, "last_layer_lr": 0.00022208724408855331, "total_loss": 15.358683586120605, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.117884635925293, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.115381240844727, "koleo_loss": -0.271484375, "ibot_loss": 5.301437854766846, "backbone_grad_norm": 0.4959676265716553, "dino_head_grad_norm": 0.02077498659491539, "ibot_head_grad_norm": 0.01596142165362835}
|
| 118 |
+
{"iteration": 1170, "iter_time": 0.3176611065864563, "data_time": 0.08805091679096222, "lr": 0.0002131065898641808, "wd": 0.33690064869943304, "mom": 0.9985977921933208, "last_layer_lr": 0.0002131065898641808, "total_loss": 15.35137939453125, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.108644485473633, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.109916687011719, "koleo_loss": -0.271484375, "ibot_loss": 5.297650337219238, "backbone_grad_norm": 0.42650794982910156, "dino_head_grad_norm": 0.020117470994591713, "ibot_head_grad_norm": 0.020476819947361946}
|
| 119 |
+
{"iteration": 1180, "iter_time": 0.275637149810791, "data_time": 0.04497503116726875, "lr": 0.0002042627193378581, "wd": 0.33956543882403073, "mom": 0.9986570097516452, "last_layer_lr": 0.0002042627193378581, "total_loss": 15.347280502319336, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.104312896728516, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.103384017944336, "koleo_loss": -0.2734375, "ibot_loss": 5.297650337219238, "backbone_grad_norm": 0.4085344672203064, "dino_head_grad_norm": 0.018715299665927887, "ibot_head_grad_norm": 0.017713215202093124}
|
| 120 |
+
{"iteration": 1190, "iter_time": 0.24994592368602753, "data_time": 0.018531858921051025, "lr": 0.00019555984171297334, "wd": 0.3421841341959295, "mom": 0.9987152029821318, "last_layer_lr": 0.00019555984171297334, "total_loss": 15.347280502319336, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.10425090789795, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.102090835571289, "koleo_loss": -0.2734375, "ibot_loss": 5.29744815826416, "backbone_grad_norm": 0.4193144142627716, "dino_head_grad_norm": 0.018909268081188202, "ibot_head_grad_norm": 0.014958771876990795}
|
| 121 |
+
{"iteration": 1200, "iter_time": 0.26037248969078064, "data_time": 0.0117651941254735, "lr": 0.00018700209908792967, "wd": 0.34475572525821446, "mom": 0.9987723494501826, "last_layer_lr": 0.00018700209908792967, "total_loss": 15.344677925109863, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.104232788085938, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.098310470581055, "koleo_loss": -0.2734375, "ibot_loss": 5.291168689727783, "backbone_grad_norm": 0.43675893545150757, "dino_head_grad_norm": 0.018826110288500786, "ibot_head_grad_norm": 0.014688564464449883}
|
| 122 |
+
{"iteration": 1210, "iter_time": 0.2564159035682678, "data_time": 0.007999991998076439, "lr": 0.00017859356448473085, "wd": 0.3472792206135786, "mom": 0.9988284271247462, "last_layer_lr": 0.00017859356448473085, "total_loss": 15.3317289352417, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.10064697265625, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.098310470581055, "koleo_loss": -0.275390625, "ibot_loss": 5.2900590896606445, "backbone_grad_norm": 0.4606385827064514, "dino_head_grad_norm": 0.018826110288500786, "ibot_head_grad_norm": 0.014490143395960331}
|
| 123 |
+
{"iteration": 1220, "iter_time": 0.29096078872680664, "data_time": 0.05840066820383072, "lr": 0.0001703382399104432, "wd": 0.3497536474065254, "mom": 0.9988834143868117, "last_layer_lr": 0.0001703382399104432, "total_loss": 15.332767486572266, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.10064697265625, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.102632522583008, "koleo_loss": -0.27734375, "ibot_loss": 5.289191722869873, "backbone_grad_norm": 0.47242051362991333, "dino_head_grad_norm": 0.02031794562935829, "ibot_head_grad_norm": 0.015522477217018604}
|
| 124 |
+
{"iteration": 1230, "iter_time": 0.3011147081851959, "data_time": 0.07022489607334137, "lr": 0.00016224005445245735, "wd": 0.3521780516984234, "mom": 0.9989372900377428, "last_layer_lr": 0.00016224005445245735, "total_loss": 15.330364227294922, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.100298881530762, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.100507736206055, "koleo_loss": -0.2734375, "ibot_loss": 5.282522201538086, "backbone_grad_norm": 0.49007850885391235, "dino_head_grad_norm": 0.02031794562935829, "ibot_head_grad_norm": 0.016958752647042274}
|
| 125 |
+
{"iteration": 1240, "iter_time": 0.28428226709365845, "data_time": 0.0540255531668663, "lr": 0.00015430286240845497, "wd": 0.3545514988352674, "mom": 0.9989900333074504, "last_layer_lr": 0.00015430286240845497, "total_loss": 15.327841758728027, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.0953369140625, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.098448753356934, "koleo_loss": -0.2734375, "ibot_loss": 5.280866622924805, "backbone_grad_norm": 0.5182631015777588, "dino_head_grad_norm": 0.017598647624254227, "ibot_head_grad_norm": 0.017202092334628105}
|
| 126 |
+
{"iteration": 1250, "iter_time": 0.2736664414405823, "data_time": 0.041699398308992386, "lr": 0.00014653044145197198, "wd": 0.3568730738080056, "mom": 0.9990416238624001, "last_layer_lr": 0.00014653044145197198, "total_loss": 15.32258129119873, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.09074592590332, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.090736389160156, "koleo_loss": -0.275390625, "ibot_loss": 5.28457498550415, "backbone_grad_norm": 0.4706212282180786, "dino_head_grad_norm": 0.017286913469433784, "ibot_head_grad_norm": 0.01808110810816288}
|
| 127 |
+
{"iteration": 1260, "iter_time": 0.24384494125843048, "data_time": 0.011545991525053978, "lr": 0.00013892649083443122, "wd": 0.3591418816052927, "mom": 0.9990920418134509, "last_layer_lr": 0.00013892649083443122, "total_loss": 15.315114974975586, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.092615127563477, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.087255477905273, "koleo_loss": -0.2734375, "ibot_loss": 5.279546737670898, "backbone_grad_norm": 0.4890872836112976, "dino_head_grad_norm": 0.02005716972053051, "ibot_head_grad_norm": 0.01656225137412548}
|
| 128 |
+
{"iteration": 1270, "iter_time": 0.23696760833263397, "data_time": 0.004995835013687611, "lr": 0.00013149462962449762, "wd": 0.36135704755853415, "mom": 0.999141267723523, "last_layer_lr": 0.00013149462962449762, "total_loss": 15.315114974975586, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.092032432556152, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.09074592590332, "koleo_loss": -0.2734375, "ibot_loss": 5.28137731552124, "backbone_grad_norm": 0.4988082945346832, "dino_head_grad_norm": 0.02014002576470375, "ibot_head_grad_norm": 0.016432207077741623}
|
| 129 |
+
{"iteration": 1280, "iter_time": 0.27325230836868286, "data_time": 0.04226334020495415, "lr": 0.0001242383949855988, "wd": 0.36351771767908725, "mom": 0.9991892826150909, "last_layer_lr": 0.0001242383949855988, "total_loss": 15.318543434143066, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.091368675231934, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.09074592590332, "koleo_loss": -0.275390625, "ibot_loss": 5.282183647155762, "backbone_grad_norm": 0.5418223142623901, "dino_head_grad_norm": 0.0199980977922678, "ibot_head_grad_norm": 0.016531409695744514}
|
| 130 |
+
{"iteration": 1290, "iter_time": 0.3125080168247223, "data_time": 0.08295029401779175, "lr": 0.00011716124049242692, "wd": 0.36562305898749053, "mom": 0.9992360679774998, "last_layer_lr": 0.00011716124049242692, "total_loss": 15.311097145080566, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.088434219360352, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.081636428833008, "koleo_loss": -0.2734375, "ibot_loss": 5.274791717529297, "backbone_grad_norm": 0.44542035460472107, "dino_head_grad_norm": 0.01916411705315113, "ibot_head_grad_norm": 0.016348212957382202}
|
| 131 |
+
{"iteration": 1300, "iter_time": 0.29785019159317017, "data_time": 0.06901589781045914, "lr": 0.00011026653448722323, "wd": 0.36767225983459245, "mom": 0.9992816057741021, "last_layer_lr": 0.00011026653448722323, "total_loss": 15.307350158691406, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.086063385009766, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.084517478942871, "koleo_loss": -0.27734375, "ibot_loss": 5.274791717529297, "backbone_grad_norm": 0.4382643699645996, "dino_head_grad_norm": 0.0201751459389925, "ibot_head_grad_norm": 0.01534013357013464}
|
| 132 |
+
{"iteration": 1310, "iter_time": 0.28154516220092773, "data_time": 0.0523165687918663, "lr": 0.00010355755847662978, "wd": 0.3696645302144581, "mom": 0.9993258784492102, "last_layer_lr": 0.00010355755847662978, "total_loss": 15.309626579284668, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.086063385009766, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.085816383361816, "koleo_loss": -0.27734375, "ibot_loss": 5.275543212890625, "backbone_grad_norm": 0.5008533596992493, "dino_head_grad_norm": 0.0201751459389925, "ibot_head_grad_norm": 0.01534013357013464}
|
| 133 |
+
{"iteration": 1320, "iter_time": 0.26771995425224304, "data_time": 0.040009476244449615, "lr": 9.703750556987118e-05, "wd": 0.3715991020689316, "mom": 0.9993688689348651, "last_layer_lr": 9.703750556987118e-05, "total_loss": 15.310250282287598, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.091201782226562, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.085102081298828, "koleo_loss": -0.27734375, "ibot_loss": 5.276608943939209, "backbone_grad_norm": 0.5252007246017456, "dino_head_grad_norm": 0.01791985146701336, "ibot_head_grad_norm": 0.015871986746788025}
|
| 134 |
+
{"iteration": 1330, "iter_time": 0.2776780128479004, "data_time": 0.05161008983850479, "lr": 9.070947895900608e-05, "wd": 0.3734752295837366, "mom": 0.9994105606574164, "last_layer_lr": 9.070947895900608e-05, "total_loss": 15.306137084960938, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.080180168151855, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.083300590515137, "koleo_loss": -0.275390625, "ibot_loss": 5.279000759124756, "backbone_grad_norm": 0.5136317014694214, "dino_head_grad_norm": 0.01791985146701336, "ibot_head_grad_norm": 0.019798455759882927}
|
| 135 |
+
{"iteration": 1340, "iter_time": 0.2998353838920593, "data_time": 0.0712476521730423, "lr": 8.457649044197615e-05, "wd": 0.37529218947600257, "mom": 0.9994509375439111, "last_layer_lr": 8.457649044197615e-05, "total_loss": 15.302677154541016, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.079534530639648, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.08254623413086, "koleo_loss": -0.275390625, "ibot_loss": 5.278166770935059, "backbone_grad_norm": 0.5776984691619873, "dino_head_grad_norm": 0.019559785723686218, "ibot_head_grad_norm": 0.02023817040026188}
|
| 136 |
+
{"iteration": 1350, "iter_time": 0.32276180386543274, "data_time": 0.07703932374715805, "lr": 7.864145898915428e-05, "wd": 0.37704928127310344, "mom": 0.9994899840282911, "last_layer_lr": 7.864145898915428e-05, "total_loss": 15.304486274719238, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.079290390014648, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.080044746398926, "koleo_loss": -0.27734375, "ibot_loss": 5.276470184326172, "backbone_grad_norm": 0.430347204208374, "dino_head_grad_norm": 0.01938391476869583, "ibot_head_grad_norm": 0.01650402694940567}
|
| 137 |
+
{"iteration": 1360, "iter_time": 0.3063186705112457, "data_time": 0.06093120574951172, "lr": 7.29072093540717e-05, "wd": 0.3787458275827039, "mom": 0.9995276850573934, "last_layer_lr": 7.29072093540717e-05, "total_loss": 15.304486274719238, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.079290390014648, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.079292297363281, "koleo_loss": -0.27734375, "ibot_loss": 5.277999401092529, "backbone_grad_norm": 0.38500428199768066, "dino_head_grad_norm": 0.0172701645642519, "ibot_head_grad_norm": 0.01650402694940567}
|
| 138 |
+
{"iteration": 1370, "iter_time": 0.25966301560401917, "data_time": 0.02852851152420044, "lr": 6.737647072898876e-05, "wd": 0.3803811743539062, "mom": 0.9995640260967534, "last_layer_lr": 6.737647072898876e-05, "total_loss": 15.297796249389648, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.079261779785156, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.082908630371094, "koleo_loss": -0.27734375, "ibot_loss": 5.276338577270508, "backbone_grad_norm": 0.4398587942123413, "dino_head_grad_norm": 0.016818325966596603, "ibot_head_grad_norm": 0.016591258347034454}
|
| 139 |
+
{"iteration": 1380, "iter_time": 0.24501001834869385, "data_time": 0.013073170557618141, "lr": 6.20518754459473e-05, "wd": 0.38195469112939984, "mom": 0.9995989931362089, "last_layer_lr": 6.20518754459473e-05, "total_loss": 15.305011749267578, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.074554443359375, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.081425666809082, "koleo_loss": -0.27734375, "ibot_loss": 5.276281356811523, "backbone_grad_norm": 0.48024117946624756, "dino_head_grad_norm": 0.02050141990184784, "ibot_head_grad_norm": 0.0175419170409441}
|
| 140 |
+
{"iteration": 1390, "iter_time": 0.24404148757457733, "data_time": 0.012390494346618652, "lr": 5.6935957723922794e-05, "wd": 0.38346577128851467, "mom": 0.9996325726953004, "last_layer_lr": 5.6935957723922794e-05, "total_loss": 15.294759750366211, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.076589584350586, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.074616432189941, "koleo_loss": -0.27734375, "ibot_loss": 5.275134086608887, "backbone_grad_norm": 0.45687368512153625, "dino_head_grad_norm": 0.01886691525578499, "ibot_head_grad_norm": 0.016347359865903854}
|
| 141 |
+
{"iteration": 1400, "iter_time": 0.27890294790267944, "data_time": 0.047981202602386475, "lr": 5.20311524626738e-05, "wd": 0.3849138322810845, "mom": 0.9996647518284686, "last_layer_lr": 5.20311524626738e-05, "total_loss": 15.292762756347656, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.077322006225586, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.075666427612305, "koleo_loss": -0.27734375, "ibot_loss": 5.273314952850342, "backbone_grad_norm": 0.45687368512153625, "dino_head_grad_norm": 0.017894655466079712, "ibot_head_grad_norm": 0.014969944022595882}
|
| 142 |
+
{"iteration": 1410, "iter_time": 0.29475289583206177, "data_time": 0.06361444294452667, "lr": 4.733979408386137e-05, "wd": 0.3862983158520316, "mom": 0.9996955181300451, "last_layer_lr": 4.733979408386137e-05, "total_loss": 15.295463562011719, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.078624725341797, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.070009231567383, "koleo_loss": -0.27734375, "ibot_loss": 5.273314952850342, "backbone_grad_norm": 0.3891368806362152, "dino_head_grad_norm": 0.018998349085450172, "ibot_head_grad_norm": 0.014849735423922539}
|
| 143 |
+
{"iteration": 1420, "iter_time": 0.26885977387428284, "data_time": 0.036152373999357224, "lr": 4.2864115419990644e-05, "wd": 0.3876186882565847, "mom": 0.9997248597390352, "last_layer_lr": 4.2864115419990644e-05, "total_loss": 15.295463562011719, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.078624725341797, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.067838668823242, "koleo_loss": -0.27734375, "ibot_loss": 5.2738847732543945, "backbone_grad_norm": 0.47476083040237427, "dino_head_grad_norm": 0.02057573013007641, "ibot_head_grad_norm": 0.015315408818423748}
|
| 144 |
+
{"iteration": 1430, "iter_time": 0.2778719365596771, "data_time": 0.044594086706638336, "lr": 3.860624665170374e-05, "wd": 0.38887444046604713, "mom": 0.99975276534369, "last_layer_lr": 3.860624665170374e-05, "total_loss": 15.297643661499023, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.079522132873535, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.071329116821289, "koleo_loss": -0.27734375, "ibot_loss": 5.272012233734131, "backbone_grad_norm": 0.4861753284931183, "dino_head_grad_norm": 0.020046930760145187, "ibot_head_grad_norm": 0.01415129005908966}
|
| 145 |
+
{"iteration": 1440, "iter_time": 0.28658896684646606, "data_time": 0.054422833025455475, "lr": 3.456821429392878e-05, "wd": 0.3900650883640381, "mom": 0.9997792241858675, "last_layer_lr": 3.456821429392878e-05, "total_loss": 15.297683715820312, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.076948165893555, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.071329116821289, "koleo_loss": -0.275390625, "ibot_loss": 5.268957138061523, "backbone_grad_norm": 0.42806121706962585, "dino_head_grad_norm": 0.018522996455430984, "ibot_head_grad_norm": 0.016578756272792816}
|
| 146 |
+
{"iteration": 1450, "iter_time": 0.2958584427833557, "data_time": 0.06392303854227066, "lr": 3.075194023136931e-05, "wd": 0.39119017293312763, "mom": 0.9998042260651806, "last_layer_lr": 3.075194023136931e-05, "total_loss": 15.292536735534668, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.075854301452637, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.080076217651367, "koleo_loss": -0.27734375, "ibot_loss": 5.268162250518799, "backbone_grad_norm": 0.46128737926483154, "dino_head_grad_norm": 0.01754402555525303, "ibot_head_grad_norm": 0.01683819480240345}
|
| 147 |
+
{"iteration": 1460, "iter_time": 0.3178738057613373, "data_time": 0.08535736799240112, "lr": 2.7159240803790108e-05, "wd": 0.3922492604317976, "mom": 0.9998277613429288, "last_layer_lr": 2.7159240803790108e-05, "total_loss": 15.291963577270508, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.07196044921875, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.07240104675293, "koleo_loss": -0.279296875, "ibot_loss": 5.2693586349487305, "backbone_grad_norm": 0.4756009578704834, "dino_head_grad_norm": 0.02254234068095684, "ibot_head_grad_norm": 0.014868808910250664}
|
| 148 |
+
{"iteration": 1470, "iter_time": 0.315117210149765, "data_time": 0.08592710644006729, "lr": 2.3791825941538785e-05, "wd": 0.3932419425616565, "mom": 0.9998498209458145, "last_layer_lr": 2.3791825941538785e-05, "total_loss": 15.287897109985352, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.072660446166992, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.070913314819336, "koleo_loss": -0.279296875, "ibot_loss": 5.269052982330322, "backbone_grad_norm": 0.40220707654953003, "dino_head_grad_norm": 0.019207822158932686, "ibot_head_grad_norm": 0.014594183303415775}
|
| 149 |
+
{"iteration": 1480, "iter_time": 0.298215389251709, "data_time": 0.07012347131967545, "lr": 2.065129835171132e-05, "wd": 0.3941678366248468, "mom": 0.9998703963694411, "last_layer_lr": 2.065129835171132e-05, "total_loss": 15.287897109985352, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.07371711730957, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.070913314819336, "koleo_loss": -0.27734375, "ibot_loss": 5.266034126281738, "backbone_grad_norm": 0.4312538206577301, "dino_head_grad_norm": 0.01809169352054596, "ibot_head_grad_norm": 0.014267580583691597}
|
| 150 |
+
{"iteration": 1490, "iter_time": 0.2585236132144928, "data_time": 0.027691924944519997, "lr": 1.773915275535004e-05, "wd": 0.3950265856715818, "mom": 0.9998894796815907, "last_layer_lr": 1.773915275535004e-05, "total_loss": 15.284089088439941, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.072273254394531, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.072824478149414, "koleo_loss": -0.27734375, "ibot_loss": 5.2662553787231445, "backbone_grad_norm": 0.5067674517631531, "dino_head_grad_norm": 0.01945098116993904, "ibot_head_grad_norm": 0.014267580583691597}
|
| 151 |
+
{"iteration": 1500, "iter_time": 0.24955205619335175, "data_time": 0.0009143591159954667, "lr": 1.5056775176037434e-05, "wd": 0.39581785863775704, "mom": 0.9999070635252835, "last_layer_lr": 1.5056775176037434e-05, "total_loss": 15.287321090698242, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.07469367980957, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.072036743164062, "koleo_loss": -0.279296875, "ibot_loss": 5.27067756652832, "backbone_grad_norm": 0.5131211280822754, "dino_head_grad_norm": 0.01945098116993904, "ibot_head_grad_norm": 0.014729749411344528}
|
| 152 |
+
{"iteration": 1510, "iter_time": 0.25062331557273865, "data_time": 0.0009368896717205644, "lr": 1.2605442280224245e-05, "wd": 0.3965413504725815, "mom": 0.999923141121613, "last_layer_lr": 1.2605442280224245e-05, "total_loss": 15.285381317138672, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.07401180267334, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.06768798828125, "koleo_loss": -0.279296875, "ibot_loss": 5.27067756652832, "backbone_grad_norm": 0.43838033080101013, "dino_head_grad_norm": 0.01738223433494568, "ibot_head_grad_norm": 0.014729749411344528}
|
| 153 |
+
{"iteration": 1520, "iter_time": 0.2337585985660553, "data_time": 0.0009155154111795127, "lr": 1.038632076960513e-05, "wd": 0.3971967822561805, "mom": 0.9999377062723596, "last_layer_lr": 1.038632076960513e-05, "total_loss": 15.284817695617676, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.070423126220703, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.071571350097656, "koleo_loss": -0.27734375, "ibot_loss": 5.268827438354492, "backbone_grad_norm": 0.3383837044239044, "dino_head_grad_norm": 0.01738223433494568, "ibot_head_grad_norm": 0.016462301835417747}
|
| 154 |
+
{"iteration": 1530, "iter_time": 0.24754440784454346, "data_time": 0.01450500451028347, "lr": 8.400466825831915e-06, "wd": 0.3977839013071248, "mom": 0.9999507533623806, "last_layer_lr": 8.400466825831915e-06, "total_loss": 15.284817695617676, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.075843811035156, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.072946548461914, "koleo_loss": -0.27734375, "ibot_loss": 5.267149925231934, "backbone_grad_norm": 0.3719562590122223, "dino_head_grad_norm": 0.017654580995440483, "ibot_head_grad_norm": 0.014931508339941502}
|
| 155 |
+
{"iteration": 1540, "iter_time": 0.288904070854187, "data_time": 0.05613002926111221, "lr": 6.648825607828386e-06, "wd": 0.3983024812798439, "mom": 0.9999622773617743, "last_layer_lr": 6.648825607828386e-06, "total_loss": 15.274763107299805, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.071630477905273, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.070683479309082, "koleo_loss": -0.27734375, "ibot_loss": 5.264135837554932, "backbone_grad_norm": 0.3719562590122223, "dino_head_grad_norm": 0.01821654476225376, "ibot_head_grad_norm": 0.013205131515860558}
|
| 156 |
+
{"iteration": 1550, "iter_time": 0.2972935736179352, "data_time": 0.06553586572408676, "lr": 5.132230801945748e-06, "wd": 0.3987523222518868, "mom": 0.9999722738278197, "last_layer_lr": 5.132230801945748e-06, "total_loss": 15.279886245727539, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.071630477905273, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.070056915283203, "koleo_loss": -0.27734375, "ibot_loss": 5.264619827270508, "backbone_grad_norm": 0.32997429370880127, "dino_head_grad_norm": 0.01815877854824066, "ibot_head_grad_norm": 0.013119593262672424}
|
| 157 |
+
{"iteration": 1560, "iter_time": 0.25594180822372437, "data_time": 0.023934494704008102, "lr": 3.851404225173012e-06, "wd": 0.39913325080099543, "mom": 0.9999807389066888, "last_layer_lr": 3.851404225173012e-06, "total_loss": 15.283597946166992, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.073049545288086, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.070491790771484, "koleo_loss": -0.27734375, "ibot_loss": 5.265931606292725, "backbone_grad_norm": 0.3865620195865631, "dino_head_grad_norm": 0.017943335697054863, "ibot_head_grad_norm": 0.014618441462516785}
|
| 158 |
+
{"iteration": 1570, "iter_time": 0.2332228720188141, "data_time": 0.0009494662517681718, "lr": 2.806955481591476e-06, "wd": 0.39944512007196303, "mom": 0.9999876693349325, "last_layer_lr": 2.806955481591476e-06, "total_loss": 15.285484313964844, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.072000503540039, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.073720932006836, "koleo_loss": -0.27734375, "ibot_loss": 5.267038345336914, "backbone_grad_norm": 0.3865620195865631, "dino_head_grad_norm": 0.017355889081954956, "ibot_head_grad_norm": 0.015177597291767597}
|
| 159 |
+
{"iteration": 1580, "iter_time": 0.2714444696903229, "data_time": 0.040499161928892136, "lr": 1.99938167223586e-06, "wd": 0.39968780983324886, "mom": 0.9999930624407388, "last_layer_lr": 1.99938167223586e-06, "total_loss": 15.285484313964844, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.073233604431152, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.074301719665527, "koleo_loss": -0.27734375, "ibot_loss": 5.267038345336914, "backbone_grad_norm": 0.43635866045951843, "dino_head_grad_norm": 0.01854509487748146, "ibot_head_grad_norm": 0.014996878802776337}
|
| 160 |
+
{"iteration": 1590, "iter_time": 0.27150797843933105, "data_time": 0.04049093648791313, "lr": 1.4290671585014682e-06, "wd": 0.39986122652333017, "mom": 0.9999969161449629, "last_layer_lr": 1.4290671585014682e-06, "total_loss": 15.287442207336426, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.072368621826172, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.073911666870117, "koleo_loss": -0.27734375, "ibot_loss": 5.269426345825195, "backbone_grad_norm": 0.4646778702735901, "dino_head_grad_norm": 0.02009003423154354, "ibot_head_grad_norm": 0.016705907881259918}
|
| 161 |
+
{"iteration": 1599, "iter_time": 0.26585331559181213, "data_time": 0.03292764350771904, "lr": 1.1188674735853883e-06, "wd": 0.3999580172602459, "mom": 0.9999990670502277, "last_layer_lr": 1.1188674735853883e-06, "total_loss": 15.281261444091797, "local_batch_size": 64.0, "global_batch_size": 64.0, "dino_local_crops_loss": 10.068781852722168, "dino_local_loss_weight": 1.0, "dino_global_crops_loss": 10.071222305297852, "koleo_loss": -0.27734375, "ibot_loss": 5.2676825523376465, "backbone_grad_norm": 0.4652988910675049, "dino_head_grad_norm": 0.018634235486388206, "ibot_head_grad_norm": 0.014847828075289726}
|