File size: 1,540 Bytes
5107f82 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 |
base_configs:
- config/base.yaml
- config/paths.yaml
name: swin_v2_tiny_1e-4_mamba_512_256
fsdp: off
fp16: off
data:
dataset: inaturalist
crop_size: 512
val_crop_size: 512
batch_size: ${train.batch_size}
val_batch_size: ${train.val_batch_size}
num_workers: 1
num_classes: 284
interpolation: bilinear
test_crop: False
aug:
auto_augment: rand-m9-mstd0.5-inc1
color_jitter: 0.4
reprob: 0.0
remode: pixel
recount: 1
mixup: 0.0
cutmix: 0.0
label_smoothing: 0.3
random_resized_crop: False
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
supercategories:
- Reptilia
model:
name: EncoderDecoderV2
num_classes: ${data.num_classes}
mlp_ratio: 4
backbone_class: swinv2_tiny_window16_256_timm
backbone:
img_size: 256
input_dim: 3
use_vanilla_backward: False
pretrained: ${oc.env:PRETRAINED_CKPT_PATH, "./ckpts"}/swinv2_tiny_patch4_window16_256.pth
upsample: False
cls_head: xl
xl_context:
skip_connection: off
enabled: off
hidden_size: 768
classification_mode: off
attention_method: mamba
tiling: naive_two_stream
n_layer: 4
resume: ''
optimizer:
name: adamw
base_lr: 1e-4
lr: 1e-4
classifier_ratio: 1.0
warmup_epochs: 0
train:
epochs: 100
batch_size: 32
val_batch_size: 1
freeze_epochs: 0
test_every: 1
test_reset: True
clip_grad: 5.0
val: False
losses:
losses:
- name: cls
type: CrossEntropy
params:
field: label
weight: 1.0
display: on
|