Is there config of Swin-L ? about swin-transformer-semantic-segmentation HOT 9 OPEN

Same issue! Please provide config for SWIN-L model

from swin-transformer-semantic-segmentation.

thank you. I love you.

from swin-transformer-semantic-segmentation.

@SwinTransformer this should be the Config of Swin-L

_base_ = [
    '../_base_/models/upernet_swin.py', '../_base_/datasets/ade20k.py',
    '../_base_/default_runtime.py', '../_base_/schedules/schedule_160k.py'
]
model = dict(
    # pretrained='https://github.com/SwinTransformer/storage/releases/download/v1.0.0/swin_large_patch4_window7_224_22k.pth',
    backbone=dict(
        embed_dim=192,
        depths=[2, 2, 18, 2],
        num_heads=[6, 12, 24, 48],
        window_size=7,
        ape=False,
        drop_path_rate=0.3,
        patch_norm=True,
        use_checkpoint=False
    ),
    decode_head=dict(
        in_channels=[192, 384, 768, 1536],
        num_classes=150
    ),
    auxiliary_head=dict(
        in_channels=768,
        num_classes=150
    ))

# AdamW optimizer, no weight decay for position embedding & layer norm in backbone
optimizer = dict(_delete_=True, type='AdamW', lr=0.00006, betas=(0.9, 0.999), weight_decay=0.01,
                 paramwise_cfg=dict(custom_keys={'absolute_pos_embed': dict(decay_mult=0.),
                                                 'relative_position_bias_table': dict(decay_mult=0.),
                                                 'norm': dict(decay_mult=0.)}))

lr_config = dict(_delete_=True, policy='poly',
                 warmup='linear',
                 warmup_iters=1500,
                 warmup_ratio=1e-6,
                 power=1.0, min_lr=0.0, by_epoch=False)

# By default, models are trained on 8 GPUs with 2 images per GPU
data=dict(samples_per_gpu=2)

from swin-transformer-semantic-segmentation.

I have the same problem about the config. However, the pretrained model are provided. You can check the link again here. Please also release the config of Swin-L with window_size=12 and it would be of great help !

from swin-transformer-semantic-segmentation.

I have the same problem about the config. However, the pretrained model are provided. You can check the link again here. Please also release the config of Swin-L with window_size=12 and it would be of great help !

Have you tried to employ the Swin-L, how is the result?

from swin-transformer-semantic-segmentation.

I have started training with Swin-B pertained model. It is failing with the multiple gpu training with some error, but is running fine with 1 gpu for me for scannet. Command: tools/dist_train.sh configs/swin/upernet_swin_base_patch4_window7_512x512_160k_scannet.py 1 --options model.pretrained=./pretrained/swin_base_patch4_window7_224_22k.pth. You could something on similar lines.

from swin-transformer-semantic-segmentation.

Same issue!

from swin-transformer-semantic-segmentation.

Hi,

I cannot find the pretrained model for Swin-L through this website stated below. Is it possible to upload this into here instead for easy reference. ('https://github.com/SwinTransformer/storage/releases/download/v1.0.0/swin_large_patch4_window7_224_22k.pth')
Thank You so much for your help.

from swin-transformer-semantic-segmentation.

hi all, does anyone know if there is a Swin-L checkpoint that has been trained on segmentation?

from swin-transformer-semantic-segmentation.