- 在 configs 目录下新增 group_vit_gcc_30e.yml 文件,配置 GroupViT 模型参数 - 更新 .gitignore 文件,忽略 .pth 文件、local_data 和 output 目录
@@ -27,6 +27,9 @@ share/python-wheels/
.installed.cfg
*.egg
MANIFEST
+*.pth
+local_data/
+output/
# PyInstaller
# Usually these files are written by a python script from a template
@@ -0,0 +1,27 @@
+_base_: 'default.yml'
+data:
+ text_aug:
+ multi_label: 3
+ batch_size: 128
+
+model:
+ type: MultiLabelContrastive
+ img_encoder:
+ type: GroupViT
+ embed_dim: 384
+ num_heads: [6, 6, 6]
+ depths: [6, 3, 3]
+ num_group_tokens: [64, 8, 0]
+ num_output_groups: [64, 8]
+ drop_rate: 0.0
+ drop_path_rate: 0.1
+ text_encoder:
+ type: TextTransformer
+ context_length: 77
+ width: 256
+ layers: 12
+ vocab_size: 49408
+ contrast_temperature: 0.07
+ proj_num_layers: 2
+ output_dim: 256
+ multi_label: ${data.text_aug.multi_label}