|
|
|
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 |
|
|