config.json

{
  "_name_or_path": "/nas/k8s/dev/mlops/chagmgang/convvit_dinov2/dinov2-remote-sensing/checkpoint-convvit-dinov1/model/checkpoint-75350",
  "architectures": [
    "ConvDINOv1"
  ],
  "base_lr": 0.002,
  "batch_size": 2048,
  "center_momentum": 0.9,
  "clip_grad": 3.0,
  "depth": [
    2,
    2,
    11
  ],
  "dino_head_n_prototypes": 65536,
  "dino_loss_weight": 1.0,
  "drop_path_rate": 0.3,
  "drop_path_uniform": false,
  "embed_dim": [
    128,
    256,
    384
  ],
  "final_momentum_teacher": 1.0,
  "global_crops_scale": [
    0.32,
    1.0
  ],
  "head_bottleneck_dim": 256,
  "head_hidden_dim": 2048,
  "head_nlayers": 3,
  "ibot_loss_weight": 1.0,
  "img_size": [
    224,
    56,
    28
  ],
  "in_chans": 3,
  "init_values": null,
  "interpolate_antialias": false,
  "interpolate_offset": 0.1,
  "koleo_loss_weight": 0.1,
  "layerwise_decay": 0.9,
  "local_crops_number": 8,
  "local_crops_scale": [
    0.05,
    0.32
  ],
  "local_crops_size": 96,
  "lr_warmup_percentile": 0.12,
  "min_lr": 1e-06,
  "mlp_ratio": [
    4.0,
    4.0,
    4.0
  ],
  "momentum_teacher": 0.992,
  "num_heads": 6,
  "num_register_tokens": 4,
  "patch_embed_lr_mult": 0.2,
  "patch_size": [
    4,
    2,
    2
  ],
  "qkv_bias": true,
  "student_temp": 0.1,
  "teacher_temp": 0.07,
  "teacher_temp_warmup_percentile": 0.32,
  "torch_dtype": "float32",
  "transformers_version": "4.41.0",
  "warmup_teacher_temp": 0.04,
  "weight_decay": 0.04,
  "weight_decay_end": 0.4
}