[Fix] fix config of maskfeat (#1424)

configs/maskfeat/maskfeat_vit-base-p16_8xb256-amp-coslr-300e_in1k.py

@@ -43,15 +43,12 @@ train_dataloader = dict(
 # model settings
 model = dict(
     type='MaskFeat',
-    data_preprocessor=dict(
-        mean=[123.675, 116.28, 103.53],
-        std=[58.395, 57.12, 57.375],
-        to_rgb=True),
     backbone=dict(type='MaskFeatViT', arch='b', patch_size=16),
     neck=dict(
         type='LinearNeck',
         in_channels=768,
         out_channels=108,
+        norm_cfg=None,
         init_cfg=dict(type='TruncNormal', layer='Linear', std=0.02, bias=0)),
     head=dict(
         type='MIMHead',
@@ -67,13 +64,13 @@ optim_wrapper = dict(
         type='AdamW', lr=2e-4 * 8, betas=(0.9, 0.999), weight_decay=0.05),
     clip_grad=dict(max_norm=0.02),
     paramwise_cfg=dict(
-        norm_decay_mult=0.0,
         bias_decay_mult=0.0,
-        # commented 'pos_embed' and 'cls_token' to avoid loss stuck situation
+        norm_decay_mult=0.0,
+        flat_decay_mult=0.0,
         custom_keys={
             # 'pos_embed': dict(decay_mult=0.),
-            'mask_token': dict(decay_mult=0.),
-            # 'cls_token': dict(decay_mult=0.)
+            # 'cls_token': dict(decay_mult=0.),
+            'mask_token': dict(decay_mult=0.)
         }))
 
 # learning rate scheduler
@@ -88,6 +85,7 @@ param_scheduler = [
     dict(
         type='CosineAnnealingLR',
         T_max=270,
+        eta_min=1e-6,
         by_epoch=True,
         begin=30,
         end=300,