File size: 6,133 Bytes
909e2c3 bf982bf 909e2c3 03dbc2c 909e2c3 bf982bf 909e2c3 bf982bf 909e2c3 bf982bf 909e2c3 bf982bf 909e2c3 03dbc2c 909e2c3 03dbc2c 909e2c3 03dbc2c 909e2c3 bf982bf 909e2c3 bf982bf 909e2c3 bf982bf 909e2c3 03dbc2c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 |
---
license: apache-2.0
base_model: google/siglip-so400m-patch14-384
tags:
- generated_from_trainer
- siglip
metrics:
- accuracy
- f1
model-index:
- name: siglip-tagger-test-3
results: []
---
# siglip-tagger-test-3
This model is a fine-tuned version of [google/siglip-so400m-patch14-384](https://huggingface.co/google/siglip-so400m-patch14-384) on an unknown dataset.
It achieves the following results on the evaluation set:
- Loss: 692.4745
- Accuracy: 0.3465
- F1: 0.9969
## Model description
This model is an experimental model that predicts danbooru tags of images.
## Example
```py
from PIL import Image
import torch
from transformers import (
AutoModelForImageClassification,
AutoImageProcessor,
)
import numpy as np
MODEL_NAME = "p1atdev/siglip-tagger-test-3"
model = AutoModelForImageClassification.from_pretrained(
MODEL_NAME, torch_dtype=torch.bfloat16, trust_remote_code=True
)
model.eval()
processor = AutoImageProcessor.from_pretrained(MODEL_NAME)
image = Image.open("sample.jpg") # load your image
inputs = processor(image, return_tensors="pt").to(model.device, model.dtype)
logits = model(**inputs).logits.detach().cpu().float()[0]
logits = np.clip(logits, 0.0, 1.0)
results = {
model.config.id2label[i]: logit for i, logit in enumerate(logits) if logit > 0
}
results = sorted(results.items(), key=lambda x: x[1], reverse=True)
for tag, score in results:
print(f"{tag}: {score*100:.2f}%")
```
## Intended uses & limitations
This model is for research use only and is not recommended for production.
Please use wd-v1-4-tagger series by SmilingWolf:
- [SmilingWolf/wd-v1-4-moat-tagger-v2](https://huggingface.co/SmilingWolf/wd-v1-4-moat-tagger-v2)
- [SmilingWolf/wd-v1-4-swinv2-tagger-v2](https://huggingface.co/SmilingWolf/wd-v1-4-swinv2-tagger-v2)
etc.
## Training and evaluation data
High quality 5000 images from danbooru. They were shuffled and split into train:eval at 4500:500. (Same as p1atdev/siglip-tagger-test-2)
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0001
- train_batch_size: 64
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 10
- num_epochs: 50
### Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 |
|:-------------:|:-----:|:----:|:---------------:|:--------:|:------:|
| 1066.981 | 1.0 | 71 | 1873.5417 | 0.1412 | 0.9939 |
| 547.3158 | 2.0 | 142 | 934.3269 | 0.1904 | 0.9964 |
| 534.6942 | 3.0 | 213 | 814.0771 | 0.2170 | 0.9966 |
| 414.1278 | 4.0 | 284 | 774.0230 | 0.2398 | 0.9967 |
| 365.4994 | 5.0 | 355 | 751.2046 | 0.2459 | 0.9967 |
| 352.3663 | 6.0 | 426 | 735.6580 | 0.2610 | 0.9967 |
| 414.3976 | 7.0 | 497 | 723.2065 | 0.2684 | 0.9968 |
| 350.8201 | 8.0 | 568 | 714.0453 | 0.2788 | 0.9968 |
| 364.5016 | 9.0 | 639 | 706.5261 | 0.2890 | 0.9968 |
| 309.1184 | 10.0 | 710 | 700.7808 | 0.2933 | 0.9968 |
| 288.5186 | 11.0 | 781 | 695.7027 | 0.3008 | 0.9968 |
| 287.4452 | 12.0 | 852 | 691.5306 | 0.3037 | 0.9968 |
| 280.9088 | 13.0 | 923 | 688.8063 | 0.3084 | 0.9969 |
| 296.8389 | 14.0 | 994 | 686.1077 | 0.3132 | 0.9968 |
| 265.1467 | 15.0 | 1065 | 683.7382 | 0.3167 | 0.9969 |
| 268.5263 | 16.0 | 1136 | 682.1683 | 0.3206 | 0.9969 |
| 309.7871 | 17.0 | 1207 | 681.1995 | 0.3199 | 0.9969 |
| 307.6475 | 18.0 | 1278 | 680.1700 | 0.3230 | 0.9969 |
| 262.0677 | 19.0 | 1349 | 679.2177 | 0.3270 | 0.9969 |
| 275.3823 | 20.0 | 1420 | 678.9730 | 0.3294 | 0.9969 |
| 273.984 | 21.0 | 1491 | 678.6031 | 0.3318 | 0.9969 |
| 273.5361 | 22.0 | 1562 | 678.1285 | 0.3332 | 0.9969 |
| 279.6474 | 23.0 | 1633 | 678.4264 | 0.3348 | 0.9969 |
| 232.5045 | 24.0 | 1704 | 678.3773 | 0.3357 | 0.9969 |
| 269.621 | 25.0 | 1775 | 678.4922 | 0.3372 | 0.9969 |
| 289.8389 | 26.0 | 1846 | 679.0094 | 0.3397 | 0.9969 |
| 256.7373 | 27.0 | 1917 | 679.5618 | 0.3407 | 0.9969 |
| 262.3969 | 28.0 | 1988 | 680.1168 | 0.3414 | 0.9969 |
| 266.2439 | 29.0 | 2059 | 681.0101 | 0.3421 | 0.9969 |
| 247.7932 | 30.0 | 2130 | 681.9800 | 0.3422 | 0.9969 |
| 246.8083 | 31.0 | 2201 | 682.8550 | 0.3416 | 0.9969 |
| 270.827 | 32.0 | 2272 | 683.9250 | 0.3434 | 0.9969 |
| 256.4384 | 33.0 | 2343 | 685.0451 | 0.3448 | 0.9969 |
| 270.461 | 34.0 | 2414 | 686.2427 | 0.3439 | 0.9969 |
| 253.8104 | 35.0 | 2485 | 687.4274 | 0.3441 | 0.9969 |
| 265.532 | 36.0 | 2556 | 688.4856 | 0.3451 | 0.9969 |
| 249.1426 | 37.0 | 2627 | 689.5027 | 0.3457 | 0.9969 |
| 229.5651 | 38.0 | 2698 | 690.4455 | 0.3455 | 0.9969 |
| 251.9008 | 39.0 | 2769 | 691.2324 | 0.3463 | 0.9969 |
| 281.8228 | 40.0 | 2840 | 691.7993 | 0.3464 | 0.9969 |
| 242.5272 | 41.0 | 2911 | 692.1788 | 0.3465 | 0.9969 |
| 229.5605 | 42.0 | 2982 | 692.3799 | 0.3465 | 0.9969 |
| 245.0876 | 43.0 | 3053 | 692.4745 | 0.3465 | 0.9969 |
| 271.22 | 44.0 | 3124 | 692.5084 | 0.3465 | 0.9969 |
| 244.3045 | 45.0 | 3195 | 692.5108 | 0.3465 | 0.9969 |
| 243.9542 | 46.0 | 3266 | 692.5128 | 0.3465 | 0.9969 |
| 274.6664 | 47.0 | 3337 | 692.5095 | 0.3465 | 0.9969 |
| 231.1361 | 48.0 | 3408 | 692.5107 | 0.3465 | 0.9969 |
| 274.5513 | 49.0 | 3479 | 692.5108 | 0.3465 | 0.9969 |
| 316.0833 | 50.0 | 3550 | 692.5107 | 0.3465 | 0.9969 |
### Framework versions
- Transformers 4.37.2
- Pytorch 2.1.2+cu118
- Datasets 2.16.1
- Tokenizers 0.15.0 |