SmilingWolf commited on
Commit
cdb0c7f
1 Parent(s): e8a7361

SwinV2 Tagger v2.1

Browse files
Files changed (6) hide show
  1. README.md +17 -2
  2. config.json +37 -0
  3. model.msgpack +3 -0
  4. model.onnx +2 -2
  5. model.safetensors +3 -0
  6. sw_jax_cv_config.json +19 -0
README.md CHANGED
@@ -16,8 +16,23 @@ Images with less than 10 general tags were filtered out.
16
  Tags with less than 600 images were filtered out.
17
 
18
  ## Validation results
19
- `P=R: threshold = 0.3771, F1 = 0.6854`
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
20
 
21
  ## Final words
22
  Subject to change and updates.
23
- Downstream users are encouraged to use tagged releases rather than relying on the head of the repo.
 
16
  Tags with less than 600 images were filtered out.
17
 
18
  ## Validation results
19
+ `v2.0: P=R: threshold = 0.3771, F1 = 0.6854`
20
+
21
+ ## What's new
22
+ Model v2.1/Dataset v2:
23
+ Re-exported to work around an ONNXRuntime v1.17.1 bug.
24
+ Bumped the minimum ONNXRuntime version to `>= 1.17.0`.
25
+ Now `timm` compatible! Load it up and give it a spin using the canonical one-liner!
26
+ Exported to `msgpack` for compatibility with the [JAX-CV](https://github.com/SmilingWolf/JAX-CV) codebase.
27
+ The batch dimension of the ONNX model is not fixed to 1 anymore. Now you can go crazy with batch inference.
28
+ No change to the trained weights themselves. There might be small prediction discrepancies across frameworks due to implementation details.
29
+
30
+ Model v2.0/Dataset v2:
31
+ Initial release.
32
+
33
+ # Runtime deps
34
+ ONNX model requires `onnxruntime >= 1.17.0`
35
 
36
  ## Final words
37
  Subject to change and updates.
38
+ Downstream users are encouraged to use tagged releases rather than relying on the head of the repo.
config.json ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architecture": "swinv2_base_window8_256",
3
+ "num_classes": 9083,
4
+ "num_features": 1024,
5
+ "global_pool": "avg",
6
+ "model_args": {
7
+ "act_layer": "gelu",
8
+ "img_size": 448,
9
+ "window_size": 14
10
+ },
11
+ "pretrained_cfg": {
12
+ "custom_load": false,
13
+ "input_size": [
14
+ 3,
15
+ 448,
16
+ 448
17
+ ],
18
+ "fixed_input_size": false,
19
+ "interpolation": "bicubic",
20
+ "crop_pct": 1.0,
21
+ "crop_mode": "center",
22
+ "mean": [
23
+ 0.5,
24
+ 0.5,
25
+ 0.5
26
+ ],
27
+ "std": [
28
+ 0.5,
29
+ 0.5,
30
+ 0.5
31
+ ],
32
+ "num_classes": 9083,
33
+ "pool_size": null,
34
+ "first_conv": null,
35
+ "classifier": null
36
+ }
37
+ }
model.msgpack ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c08431fe95c5ce2d0cdfd16e42e78763799f620c09ad51d16c141ffd67083d9c
3
+ size 406487497
model.onnx CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:67740df7ede9a53e50d6e29c6a5c0d6c862f1876c22545d810515bad3ae17bb1
3
- size 455481275
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:04ec04fdf7db74b4fed7f4b52f52e04dec4dbad9e4d88d2d178f334079a29fde
3
+ size 455409152
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d70a9c7138318dbdc487fb9bcaa1b86ae9a41f6132d823a257e8d4a5fb42636
3
+ size 384859388
sw_jax_cv_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "image_size": 448,
3
+ "model_name": "swinv2_base",
4
+ "model_args": {
5
+ "image_size": 448,
6
+ "patch_size": 4,
7
+ "in_chans": 3,
8
+ "num_classes": 9083,
9
+ "embed_dim": 128,
10
+ "window_size": 14,
11
+ "mlp_ratio": 4.0,
12
+ "qkv_bias": true,
13
+ "drop_rate": 0.0,
14
+ "attn_drop_rate": 0.0,
15
+ "drop_path_rate": 0.1,
16
+ "patch_norm": true,
17
+ "layer_norm_eps": 1e-05
18
+ }
19
+ }