Hugging Face's logo

timm
/
vit_dwee_patch16_reg1_gap_256.sbb_nadamuon_in1k

Image Classification
timm
Safetensors
Transformers
Model card Files
xet
 Community
rwightman commited on
Commit
61a22c5
·
verified ·
1 Parent(s): 99e7a1f

Update model config and README

Browse files
Files changed (2) hide show
  1. README.md +7 -5
  2. config.json +1 -1
README.md CHANGED
@@ -9,7 +9,7 @@ license: apache-2.0
9
  datasets:
10
  - imagenet-1k
11
  ---
12
- # Model card for vit_dwee_patch16_reg1_gap_256.sbb_adamuon_in1k
13
 
14
  A Vision Transformer (ViT) image classification model. This is a `timm` specific variation of the architecture with registers, global average pooling, differential attention.
15
 
@@ -23,7 +23,8 @@ There are a number of models in the lower end of model scales that originate in
23
  | pwee | 256 | 1280 (5) | 4 | 16 (parallel) | y |
24
  | dpwee | 256 | 1280 (5) | 4 | 16 (parallel + differential) | y |
25
  | small | 384 | 1536 (4) | 6 | 12 | n |
26
- | little | 320 | 1792 (5.6) | 5 | 14 | y |
 
27
  | medium | 512 | 2048 (4) | 8 | 12 | y |
28
  | mediumd | 512 | 2048 (4) | 8 | 20 | y |
29
  | betwixt | 640 | 2560 (4) | 10 | 12 | y |
@@ -63,7 +64,7 @@ img = Image.open(urlopen(
63
  'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/beignets-task-guide.png'
64
  ))
65
 
66
- model = timm.create_model('vit_dwee_patch16_reg1_gap_256.sbb_adamuon_in1k', pretrained=True)
67
  model = model.eval()
68
 
69
  # get model specific transforms (normalization, resize)
@@ -86,7 +87,7 @@ img = Image.open(urlopen(
86
  ))
87
 
88
  model = timm.create_model(
89
- 'vit_dwee_patch16_reg1_gap_256.sbb_adamuon_in1k',
90
  pretrained=True,
91
  features_only=True,
92
  )
@@ -119,7 +120,7 @@ img = Image.open(urlopen(
119
  ))
120
 
121
  model = timm.create_model(
122
- 'vit_dwee_patch16_reg1_gap_256.sbb_adamuon_in1k',
123
  pretrained=True,
124
  num_classes=0, # remove classifier nn.Linear
125
  )
@@ -162,6 +163,7 @@ output = model.forward_head(output, pre_logits=True)
162
  | [vit_betwixt_patch16_reg1_gap_256.sbb_in1k](https://huggingface.co/timm/vit_betwixt_patch16_reg1_gap_256.sbb_in1k) | 83.628 | 96.544 | 60.4 | 256 |
163
  | [vit_medium_patch16_reg4_gap_256.sbb_in1k](https://huggingface.co/timm/vit_medium_patch16_reg4_gap_256.sbb_in1k) | 83.47 | 96.622 | 38.88 | 256 |
164
  | [vit_medium_patch16_reg1_gap_256.sbb_in1k](https://huggingface.co/timm/vit_medium_patch16_reg1_gap_256.sbb_in1k) | 83.462 | 96.548 | 38.88 | 256 |
 
165
  | [vit_little_patch16_reg4_gap_256.sbb_in1k](https://huggingface.co/timm/vit_little_patch16_reg4_gap_256.sbb_in1k) | 82.514 | 96.262 | 22.52 | 256 |
166
  | [vit_dwee_patch16_reg1_gap_256.sbb_nadamuon_in1k](https://huggingface.co/timm/vit_dwee_patch16_reg1_gap_256.sbb_nadamuon_in1k) | 81.808 | 95.892 | 13.43 | 256 |
167
  | [vit_dpwee_patch16_reg1_gap_256.sbb_nadamuon_in1k](https://huggingface.co/timm/vit_dpwee_patch16_reg1_gap_256.sbb_nadamuon_in1k) | 81.660 | 95.808 | 15.25 | 256 |
 
9
  datasets:
10
  - imagenet-1k
11
  ---
12
+ # Model card for vit_dwee_patch16_reg1_gap_256.sbb_nadamuon_in1k
13
 
14
  A Vision Transformer (ViT) image classification model. This is a `timm` specific variation of the architecture with registers, global average pooling, differential attention.
15
 
 
23
  | pwee | 256 | 1280 (5) | 4 | 16 (parallel) | y |
24
  | dpwee | 256 | 1280 (5) | 4 | 16 (parallel + differential) | y |
25
  | small | 384 | 1536 (4) | 6 | 12 | n |
26
+ | little | 320 | 1792 (5.6) | 5 | 14 (differential) | y |
27
+ | dlittle | 320 | 1792 (5.6) | 5 | 14 | y |
28
  | medium | 512 | 2048 (4) | 8 | 12 | y |
29
  | mediumd | 512 | 2048 (4) | 8 | 20 | y |
30
  | betwixt | 640 | 2560 (4) | 10 | 12 | y |
 
64
  'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/beignets-task-guide.png'
65
  ))
66
 
67
+ model = timm.create_model('vit_dwee_patch16_reg1_gap_256.sbb_nadamuon_in1k', pretrained=True)
68
  model = model.eval()
69
 
70
  # get model specific transforms (normalization, resize)
 
87
  ))
88
 
89
  model = timm.create_model(
90
+ 'vit_dwee_patch16_reg1_gap_256.sbb_nadamuon_in1k',
91
  pretrained=True,
92
  features_only=True,
93
  )
 
120
  ))
121
 
122
  model = timm.create_model(
123
+ 'vit_dwee_patch16_reg1_gap_256.sbb_nadamuon_in1k',
124
  pretrained=True,
125
  num_classes=0, # remove classifier nn.Linear
126
  )
 
163
  | [vit_betwixt_patch16_reg1_gap_256.sbb_in1k](https://huggingface.co/timm/vit_betwixt_patch16_reg1_gap_256.sbb_in1k) | 83.628 | 96.544 | 60.4 | 256 |
164
  | [vit_medium_patch16_reg4_gap_256.sbb_in1k](https://huggingface.co/timm/vit_medium_patch16_reg4_gap_256.sbb_in1k) | 83.47 | 96.622 | 38.88 | 256 |
165
  | [vit_medium_patch16_reg1_gap_256.sbb_in1k](https://huggingface.co/timm/vit_medium_patch16_reg1_gap_256.sbb_in1k) | 83.462 | 96.548 | 38.88 | 256 |
166
+ | [vit_dlittle_patch16_reg1_gap_256.sbb_in1k](https://huggingface.co/timm/vit_dlittle_patch16_reg1_gap_256.sbb_in1k) | 83.212 | 96.358 | 22.52 | 256 |
167
  | [vit_little_patch16_reg4_gap_256.sbb_in1k](https://huggingface.co/timm/vit_little_patch16_reg4_gap_256.sbb_in1k) | 82.514 | 96.262 | 22.52 | 256 |
168
  | [vit_dwee_patch16_reg1_gap_256.sbb_nadamuon_in1k](https://huggingface.co/timm/vit_dwee_patch16_reg1_gap_256.sbb_nadamuon_in1k) | 81.808 | 95.892 | 13.43 | 256 |
169
  | [vit_dpwee_patch16_reg1_gap_256.sbb_nadamuon_in1k](https://huggingface.co/timm/vit_dpwee_patch16_reg1_gap_256.sbb_nadamuon_in1k) | 81.660 | 95.808 | 15.25 | 256 |
config.json CHANGED
@@ -4,7 +4,7 @@
4
  "num_features": 256,
5
  "global_pool": "avg",
6
  "pretrained_cfg": {
7
- "tag": "sbb_adamuon_in1k",
8
  "custom_load": false,
9
  "input_size": [
10
  3,
 
4
  "num_features": 256,
5
  "global_pool": "avg",
6
  "pretrained_cfg": {
7
+ "tag": "sbb_nadamuon_in1k",
8
  "custom_load": false,
9
  "input_size": [
10
  3,