Hugging Face's logo

zeromodels
/
swin_base_patch4_window12_384_ms_in22k_ft_in1k

Image Classification
ZeroModels
Keras
PyTorch
JAX
TensorFlow
swin
backbone
Model card Files
xet
 Community
IMvision12 commited on
Commit
44aca9a
·
verified ·
1 Parent(s): 9a4cf6c

Migrate to zeromodels (rename kf_*.json -> zm_*.json, fix refs in config + README, ensure tag + badge)

Browse files
Files changed (2) hide show
  1. README.md +27 -27
  2. kf_config.json → zm_config.json +29 -29
README.md CHANGED
@@ -2,10 +2,10 @@
2
  pipeline_tag: image-classification
3
  license: mit
4
  base_model: timm/swin_base_patch4_window12_384.ms_in22k_ft_in1k
5
- library_name: kerasformers
6
  tags:
7
  - keras
8
- - kerasformers
9
  - image-classification
10
  - swin
11
  - backbone
@@ -15,13 +15,13 @@ tags:
15
  - tf
16
  ---
17
 
18
- ## ***See [our collection](https://huggingface.co/collections/kerasformers/swin-transformer-6a6c7c6d86b6537929511843) for all versions of Swin Transformer.***
19
 
20
  # Run Swin Transformer with Keras 3: JAX, PyTorch, or TensorFlow
21
 
22
- [![GitHub](https://img.shields.io/badge/GitHub-KerasFormers-black?logo=github)](https://github.com/IMvision12/KerasFormers) [![Docs](https://img.shields.io/badge/Docs-Swin-blue)](https://imvision12.github.io/KerasFormers/classification_backbones/) [![Collection](https://img.shields.io/badge/HF-Swin%20collection-yellow)](https://huggingface.co/collections/kerasformers/swin-transformer-6a6c7c6d86b6537929511843)
23
 
24
- # kerasformers/swin_base_patch4_window12_384_ms_in22k_ft_in1k
25
 
26
  Paper: [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows (arXiv:2103.14030)](https://arxiv.org/abs/2103.14030) · [HF Papers](https://huggingface.co/papers/2103.14030)
27
 
@@ -29,7 +29,7 @@ Swin Transformer builds hierarchical feature maps with shifted-window attention.
29
 
30
  For more details on the model, please go to the upstream [model card](https://huggingface.co/timm/swin_base_patch4_window12_384.ms_in22k_ft_in1k).
31
 
32
- Pure-**Keras 3** conversion of [`timm/swin_base_patch4_window12_384.ms_in22k_ft_in1k`](https://huggingface.co/timm/swin_base_patch4_window12_384.ms_in22k_ft_in1k) for [kerasformers](https://github.com/IMvision12/KerasFormers). One implementation runs unmodified on **TensorFlow / Torch / JAX**.
33
 
34
  This is an **image-classification / backbone** checkpoint (`SwinImageClassify` / `SwinModel`).
35
 
@@ -41,11 +41,11 @@ os.environ["KERAS_BACKEND"] = "torch" # or "jax" / "tensorflow"
41
 
42
  from PIL import Image
43
  import numpy as np
44
- from kerasformers.models.swin import SwinImageClassify, SwinModel
45
 
46
- model = SwinImageClassify.from_weights("kerasformers/swin_base_patch4_window12_384_ms_in22k_ft_in1k")
47
  backbone = SwinModel.from_weights(
48
- "kerasformers/swin_base_patch4_window12_384_ms_in22k_ft_in1k", as_backbone=True
49
  )
50
 
51
  image = Image.open("your_image.jpg").convert("RGB")
@@ -56,31 +56,31 @@ feats = backbone(x)
56
  print(len(feats), [tuple(f.shape) for f in feats])
57
  ```
58
 
59
- Load any Swin Transformer variant the same way with `from_weights("kerasformers/<variant>")`:
60
 
61
  | Variant | Hub |
62
  |---|---|
63
- | `swin_base_patch4_window12_384_ms_in1k` | [`kerasformers/swin_base_patch4_window12_384_ms_in1k`](https://huggingface.co/kerasformers/swin_base_patch4_window12_384_ms_in1k) |
64
- | `swin_base_patch4_window12_384_ms_in22k` | [`kerasformers/swin_base_patch4_window12_384_ms_in22k`](https://huggingface.co/kerasformers/swin_base_patch4_window12_384_ms_in22k) |
65
- | `swin_base_patch4_window12_384_ms_in22k_ft_in1k` | [`kerasformers/swin_base_patch4_window12_384_ms_in22k_ft_in1k`](https://huggingface.co/kerasformers/swin_base_patch4_window12_384_ms_in22k_ft_in1k) |
66
- | `swin_base_patch4_window7_224_ms_in1k` | [`kerasformers/swin_base_patch4_window7_224_ms_in1k`](https://huggingface.co/kerasformers/swin_base_patch4_window7_224_ms_in1k) |
67
- | `swin_base_patch4_window7_224_ms_in22k` | [`kerasformers/swin_base_patch4_window7_224_ms_in22k`](https://huggingface.co/kerasformers/swin_base_patch4_window7_224_ms_in22k) |
68
- | `swin_base_patch4_window7_224_ms_in22k_ft_in1k` | [`kerasformers/swin_base_patch4_window7_224_ms_in22k_ft_in1k`](https://huggingface.co/kerasformers/swin_base_patch4_window7_224_ms_in22k_ft_in1k) |
69
- | `swin_large_patch4_window12_384_ms_in22k` | [`kerasformers/swin_large_patch4_window12_384_ms_in22k`](https://huggingface.co/kerasformers/swin_large_patch4_window12_384_ms_in22k) |
70
- | `swin_large_patch4_window12_384_ms_in22k_ft_in1k` | [`kerasformers/swin_large_patch4_window12_384_ms_in22k_ft_in1k`](https://huggingface.co/kerasformers/swin_large_patch4_window12_384_ms_in22k_ft_in1k) |
71
- | `swin_large_patch4_window7_224_ms_in22k` | [`kerasformers/swin_large_patch4_window7_224_ms_in22k`](https://huggingface.co/kerasformers/swin_large_patch4_window7_224_ms_in22k) |
72
- | `swin_large_patch4_window7_224_ms_in22k_ft_in1k` | [`kerasformers/swin_large_patch4_window7_224_ms_in22k_ft_in1k`](https://huggingface.co/kerasformers/swin_large_patch4_window7_224_ms_in22k_ft_in1k) |
73
- | `swin_small_patch4_window7_224_ms_in1k` | [`kerasformers/swin_small_patch4_window7_224_ms_in1k`](https://huggingface.co/kerasformers/swin_small_patch4_window7_224_ms_in1k) |
74
- | `swin_small_patch4_window7_224_ms_in22k` | [`kerasformers/swin_small_patch4_window7_224_ms_in22k`](https://huggingface.co/kerasformers/swin_small_patch4_window7_224_ms_in22k) |
75
- | `swin_small_patch4_window7_224_ms_in22k_ft_in1k` | [`kerasformers/swin_small_patch4_window7_224_ms_in22k_ft_in1k`](https://huggingface.co/kerasformers/swin_small_patch4_window7_224_ms_in22k_ft_in1k) |
76
- | `swin_tiny_patch4_window7_224_ms_in1k` | [`kerasformers/swin_tiny_patch4_window7_224_ms_in1k`](https://huggingface.co/kerasformers/swin_tiny_patch4_window7_224_ms_in1k) |
77
- | `swin_tiny_patch4_window7_224_ms_in22k` | [`kerasformers/swin_tiny_patch4_window7_224_ms_in22k`](https://huggingface.co/kerasformers/swin_tiny_patch4_window7_224_ms_in22k) |
78
 
79
  ## Tips
80
 
81
- - Set `KERAS_BACKEND` **before** importing Keras / kerasformers.
82
  - `SwinImageClassify` returns class logits; `SwinModel` returns features (`as_backbone=True` for multi-scale stages).
83
- - See [docs](https://imvision12.github.io/KerasFormers/classification_backbones/) and [Loading Weights](https://imvision12.github.io/KerasFormers/loading_weights/).
84
  - Upstream / timm checkpoints: `SwinImageClassify.from_weights("hf:timm/swin_base_patch4_window12_384.ms_in22k_ft_in1k")`.
85
 
86
  ## Special Thanks
 
2
  pipeline_tag: image-classification
3
  license: mit
4
  base_model: timm/swin_base_patch4_window12_384.ms_in22k_ft_in1k
5
+ library_name: zeromodels
6
  tags:
7
  - keras
8
+ - zeromodels
9
  - image-classification
10
  - swin
11
  - backbone
 
15
  - tf
16
  ---
17
 
18
+ ## ***See [our collection](https://huggingface.co/collections/zeromodels/swin-transformer-6a6c7c6d86b6537929511843) for all versions of Swin Transformer.***
19
 
20
  # Run Swin Transformer with Keras 3: JAX, PyTorch, or TensorFlow
21
 
22
+ [![GitHub](https://img.shields.io/badge/GitHub-ZeroModels-black?logo=github)](https://github.com/IMvision12/ZeroModels) [![Docs](https://img.shields.io/badge/Docs-Swin-blue)](https://imvision12.github.io/ZeroModels/classification_backbones/) [![Collection](https://img.shields.io/badge/HF-Swin%20collection-yellow)](https://huggingface.co/collections/zeromodels/swin-transformer-6a6c7c6d86b6537929511843)
23
 
24
+ # zeromodels/swin_base_patch4_window12_384_ms_in22k_ft_in1k
25
 
26
  Paper: [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows (arXiv:2103.14030)](https://arxiv.org/abs/2103.14030) · [HF Papers](https://huggingface.co/papers/2103.14030)
27
 
 
29
 
30
  For more details on the model, please go to the upstream [model card](https://huggingface.co/timm/swin_base_patch4_window12_384.ms_in22k_ft_in1k).
31
 
32
+ Pure-**Keras 3** conversion of [`timm/swin_base_patch4_window12_384.ms_in22k_ft_in1k`](https://huggingface.co/timm/swin_base_patch4_window12_384.ms_in22k_ft_in1k) for [zeromodels](https://github.com/IMvision12/ZeroModels). One implementation runs unmodified on **TensorFlow / Torch / JAX**.
33
 
34
  This is an **image-classification / backbone** checkpoint (`SwinImageClassify` / `SwinModel`).
35
 
 
41
 
42
  from PIL import Image
43
  import numpy as np
44
+ from zeromodels.models.swin import SwinImageClassify, SwinModel
45
 
46
+ model = SwinImageClassify.from_weights("zeromodels/swin_base_patch4_window12_384_ms_in22k_ft_in1k")
47
  backbone = SwinModel.from_weights(
48
+ "zeromodels/swin_base_patch4_window12_384_ms_in22k_ft_in1k", as_backbone=True
49
  )
50
 
51
  image = Image.open("your_image.jpg").convert("RGB")
 
56
  print(len(feats), [tuple(f.shape) for f in feats])
57
  ```
58
 
59
+ Load any Swin Transformer variant the same way with `from_weights("zeromodels/<variant>")`:
60
 
61
  | Variant | Hub |
62
  |---|---|
63
+ | `swin_base_patch4_window12_384_ms_in1k` | [`zeromodels/swin_base_patch4_window12_384_ms_in1k`](https://huggingface.co/zeromodels/swin_base_patch4_window12_384_ms_in1k) |
64
+ | `swin_base_patch4_window12_384_ms_in22k` | [`zeromodels/swin_base_patch4_window12_384_ms_in22k`](https://huggingface.co/zeromodels/swin_base_patch4_window12_384_ms_in22k) |
65
+ | `swin_base_patch4_window12_384_ms_in22k_ft_in1k` | [`zeromodels/swin_base_patch4_window12_384_ms_in22k_ft_in1k`](https://huggingface.co/zeromodels/swin_base_patch4_window12_384_ms_in22k_ft_in1k) |
66
+ | `swin_base_patch4_window7_224_ms_in1k` | [`zeromodels/swin_base_patch4_window7_224_ms_in1k`](https://huggingface.co/zeromodels/swin_base_patch4_window7_224_ms_in1k) |
67
+ | `swin_base_patch4_window7_224_ms_in22k` | [`zeromodels/swin_base_patch4_window7_224_ms_in22k`](https://huggingface.co/zeromodels/swin_base_patch4_window7_224_ms_in22k) |
68
+ | `swin_base_patch4_window7_224_ms_in22k_ft_in1k` | [`zeromodels/swin_base_patch4_window7_224_ms_in22k_ft_in1k`](https://huggingface.co/zeromodels/swin_base_patch4_window7_224_ms_in22k_ft_in1k) |
69
+ | `swin_large_patch4_window12_384_ms_in22k` | [`zeromodels/swin_large_patch4_window12_384_ms_in22k`](https://huggingface.co/zeromodels/swin_large_patch4_window12_384_ms_in22k) |
70
+ | `swin_large_patch4_window12_384_ms_in22k_ft_in1k` | [`zeromodels/swin_large_patch4_window12_384_ms_in22k_ft_in1k`](https://huggingface.co/zeromodels/swin_large_patch4_window12_384_ms_in22k_ft_in1k) |
71
+ | `swin_large_patch4_window7_224_ms_in22k` | [`zeromodels/swin_large_patch4_window7_224_ms_in22k`](https://huggingface.co/zeromodels/swin_large_patch4_window7_224_ms_in22k) |
72
+ | `swin_large_patch4_window7_224_ms_in22k_ft_in1k` | [`zeromodels/swin_large_patch4_window7_224_ms_in22k_ft_in1k`](https://huggingface.co/zeromodels/swin_large_patch4_window7_224_ms_in22k_ft_in1k) |
73
+ | `swin_small_patch4_window7_224_ms_in1k` | [`zeromodels/swin_small_patch4_window7_224_ms_in1k`](https://huggingface.co/zeromodels/swin_small_patch4_window7_224_ms_in1k) |
74
+ | `swin_small_patch4_window7_224_ms_in22k` | [`zeromodels/swin_small_patch4_window7_224_ms_in22k`](https://huggingface.co/zeromodels/swin_small_patch4_window7_224_ms_in22k) |
75
+ | `swin_small_patch4_window7_224_ms_in22k_ft_in1k` | [`zeromodels/swin_small_patch4_window7_224_ms_in22k_ft_in1k`](https://huggingface.co/zeromodels/swin_small_patch4_window7_224_ms_in22k_ft_in1k) |
76
+ | `swin_tiny_patch4_window7_224_ms_in1k` | [`zeromodels/swin_tiny_patch4_window7_224_ms_in1k`](https://huggingface.co/zeromodels/swin_tiny_patch4_window7_224_ms_in1k) |
77
+ | `swin_tiny_patch4_window7_224_ms_in22k` | [`zeromodels/swin_tiny_patch4_window7_224_ms_in22k`](https://huggingface.co/zeromodels/swin_tiny_patch4_window7_224_ms_in22k) |
78
 
79
  ## Tips
80
 
81
+ - Set `KERAS_BACKEND` **before** importing Keras / zeromodels.
82
  - `SwinImageClassify` returns class logits; `SwinModel` returns features (`as_backbone=True` for multi-scale stages).
83
+ - See [docs](https://imvision12.github.io/ZeroModels/classification_backbones/) and [Loading Weights](https://imvision12.github.io/ZeroModels/loading_weights/).
84
  - Upstream / timm checkpoints: `SwinImageClassify.from_weights("hf:timm/swin_base_patch4_window12_384.ms_in22k_ft_in1k")`.
85
 
86
  ## Special Thanks
kf_config.json → zm_config.json RENAMED
@@ -1,30 +1,30 @@
1
- {
2
- "library_name": "kerasformers",
3
- "kerasformers_version": "1.2.1",
4
- "model_module": "kerasformers.models.swin",
5
- "model_class": "SwinImageClassify",
6
- "variant": "swin_base_patch4_window12_384_ms_in22k_ft_in1k",
7
- "weights": "model.weights.h5",
8
- "schema_version": 2,
9
- "weight_dtype": "float32",
10
- "model_type": "swin",
11
- "vision_config": {
12
- "window_size": 12,
13
- "embed_dim": 128,
14
- "depths": [
15
- 2,
16
- 2,
17
- 18,
18
- 2
19
- ],
20
- "num_heads": [
21
- 4,
22
- 8,
23
- 16,
24
- 32
25
- ],
26
- "pretrain_size": 384,
27
- "image_size": 384,
28
- "num_classes": 1000
29
- }
30
  }
 
1
+ {
2
+ "library_name": "zeromodels",
3
+ "zeromodels_version": "1.2.1",
4
+ "model_module": "zeromodels.models.swin",
5
+ "model_class": "SwinImageClassify",
6
+ "variant": "swin_base_patch4_window12_384_ms_in22k_ft_in1k",
7
+ "weights": "model.weights.h5",
8
+ "schema_version": 2,
9
+ "weight_dtype": "float32",
10
+ "model_type": "swin",
11
+ "vision_config": {
12
+ "window_size": 12,
13
+ "embed_dim": 128,
14
+ "depths": [
15
+ 2,
16
+ 2,
17
+ 18,
18
+ 2
19
+ ],
20
+ "num_heads": [
21
+ 4,
22
+ 8,
23
+ 16,
24
+ 32
25
+ ],
26
+ "pretrain_size": 384,
27
+ "image_size": 384,
28
+ "num_classes": 1000
29
+ }
30
  }