Upload models/model_configs/fcn_sere-small_finetuned_fp16_8x32_224x224_3600_imagenets919.py with huggingface_hub

models/model_configs/fcn_sere-small_finetuned_fp16_8x32_224x224_3600_imagenets919.py

@@ -0,0 +1,107 @@
+# _base_ = [
+#     '../_base_/models/fcn_r50-d8.py', '../_base_/datasets/imagenets.py',
+#     '../_base_/default_runtime.py', '../_base_/schedules/schedule_20k.py'
+# ]
+
+
+
+model = dict(
+    type='EncoderDecoder',  # Add this line to specify the model type
+    pretrained='models/checkpoints/sere_finetuned_vit_small_ep100.pth',
+    backbone=dict(
+        type='VisionTransformer',
+        img_size=(224, 224),
+        patch_size=16,
+        in_channels=3,
+        embed_dims=384,
+        num_layers=12,
+        num_heads=6,
+        mlp_ratio=4,
+        out_indices=-1, #(2, 5, 8, 11)
+        qkv_bias=True,
+        drop_rate=0.0,
+        attn_drop_rate=0.0,
+        drop_path_rate=0.1,
+        with_cls_token=True,
+        norm_cfg=dict(type='LN', eps=1e-6),
+        act_cfg=dict(type='GELU'),
+        norm_eval=False,
+        final_norm=True,
+        interpolate_mode='bicubic'),
+    decode_head=dict(
+        type='FCNHead',  # Add this line to specify the decode_head type
+        in_channels=384,
+        channels=384,
+        num_convs=0,
+        dropout_ratio=0.0,
+        num_classes=920,
+        ignore_index=1000,
+        # downsample_label_ratio=8,
+        init_cfg=dict(
+            type='TruncNormal', std=2e-5, override=dict(name='conv_seg'))),
+    auxiliary_head=None)
+
+# model = dict(
+#     pretrained='../checkpoints/sere_finetuned_vit_small_ep100.pth',
+#     backbone=dict(
+#         # _delete_=True,
+#         type='VisionTransformer',
+#         img_size=(224, 224),
+#         patch_size=16,
+#         in_channels=3,
+#         embed_dims=384,
+#         num_layers=12,
+#         num_heads=6,
+#         mlp_ratio=4,
+#         out_indices=(2, 5, 8, 11),
+#         qkv_bias=True,
+#         drop_rate=0.0,
+#         attn_drop_rate=0.0,
+#         drop_path_rate=0.1,
+#         with_cls_token=True,
+#         norm_cfg=dict(type='LN', eps=1e-6),
+#         act_cfg=dict(type='GELU'),
+#         norm_eval=False,
+#         final_norm=True,
+#         interpolate_mode='bicubic'),
+#     decode_head=dict(
+#         in_channels=384,
+#         channels=384,
+#         num_convs=0,
+#         dropout_ratio=0.0,
+#         num_classes=920,
+#         ignore_index=1000,
+#         downsample_label_ratio=8,
+#         init_cfg=dict(
+#             type='TruncNormal', std=2e-5, override=dict(name='conv_seg'))),
+#     auxiliary_head=None)
+
+optimizer = dict(
+    _delete_=True,
+    constructor='LearningRateDecayOptimizerConstructor',
+    type='AdamW',
+    lr=5e-4,
+    betas=(0.9, 0.999),
+    weight_decay=0.05,
+    paramwise_cfg=dict(
+        num_layers=12, decay_rate=0.50, decay_type='layer_wise'))
+
+lr_config = dict(
+    _delete_=True,
+    policy='CosineAnnealing',
+    warmup='linear',
+    warmup_iters=180,
+    warmup_ratio=1e-6,
+    min_lr=1e-6,
+    by_epoch=False)
+
+# mixed precision
+fp16 = dict(loss_scale='dynamic')
+
+# By default, models are trained on 8 GPUs with 32 images per GPU
+data = dict(samples_per_gpu=32)
+
+# runtime settings
+runner = dict(type='IterBasedRunner', max_iters=3600)
+checkpoint_config = dict(by_epoch=False, interval=3600)
+evaluation = dict(interval=360, metric='mIoU', pre_eval=True)