nielsr HF staff commited on
Commit
d430295
1 Parent(s): 61d034f

Update README

Browse files
Files changed (1) hide show
  1. README.md +6 -6
README.md CHANGED
@@ -16,6 +16,8 @@ The Vision Transformer (ViT) is a transformer encoder model (BERT-like) pretrain
16
 
17
  Images are presented to the model as a sequence of fixed-size patches (resolution 16x16), which are linearly embedded. One also adds a [CLS] token to the beginning of a sequence to use it for classification tasks. One also adds absolute position embeddings before feeding the sequence to the layers of the Transformer encoder.
18
 
 
 
19
  By pre-training the model, it learns an inner representation of images that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled images for instance, you can train a standard classifier by placing a linear layer on top of the pre-trained encoder. One typically places a linear layer on top of the [CLS] token, as the last hidden state of this token can be seen as a representation of an entire image.
20
 
21
  ## Intended uses & limitations
@@ -28,18 +30,16 @@ fine-tuned versions on a task that interests you.
28
  Here is how to use this model to classify an image of the COCO 2017 dataset into one of the 21k ImageNet-21k classes:
29
 
30
  ```python
31
- from transformers import ViTFeatureExtractor, ViTForImageClassification
32
  from PIL import Image
33
  import requests
34
  url = 'http://images.cocodataset.org/val2017/000000039769.jpg'
35
  image = Image.open(requests.get(url, stream=True).raw)
36
  feature_extractor = ViTFeatureExtractor.from_pretrained('google/vit-base-patch16-224-in21k')
37
- model = ViTForImageClassification.from_pretrained('google/vit-base-patch16-224-in21k')
38
- inputs = feature_extractor(images=image)
39
  outputs = model(**inputs)
40
- logits = outputs.logits
41
- # model predicts one of the 21k ImageNet-21k classes
42
- predicted_class = logits.argmax(-1).item()
43
  ```
44
 
45
  Currently, both the feature extractor and model support PyTorch. Tensorflow and JAX/FLAX are coming soon, and the API of ViTFeatureExtractor might change.
 
16
 
17
  Images are presented to the model as a sequence of fixed-size patches (resolution 16x16), which are linearly embedded. One also adds a [CLS] token to the beginning of a sequence to use it for classification tasks. One also adds absolute position embeddings before feeding the sequence to the layers of the Transformer encoder.
18
 
19
+ Note that this model does not provide any fine-tuned heads, as these were zero'd by Google researchers. However, the model does include the pre-trained pooler, which can be used for downstream tasks (i.e. such as image classification).
20
+
21
  By pre-training the model, it learns an inner representation of images that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled images for instance, you can train a standard classifier by placing a linear layer on top of the pre-trained encoder. One typically places a linear layer on top of the [CLS] token, as the last hidden state of this token can be seen as a representation of an entire image.
22
 
23
  ## Intended uses & limitations
 
30
  Here is how to use this model to classify an image of the COCO 2017 dataset into one of the 21k ImageNet-21k classes:
31
 
32
  ```python
33
+ from transformers import ViTFeatureExtractor, ViTModel
34
  from PIL import Image
35
  import requests
36
  url = 'http://images.cocodataset.org/val2017/000000039769.jpg'
37
  image = Image.open(requests.get(url, stream=True).raw)
38
  feature_extractor = ViTFeatureExtractor.from_pretrained('google/vit-base-patch16-224-in21k')
39
+ model = ViTModel.from_pretrained('google/vit-base-patch16-224-in21k')
40
+ inputs = feature_extractor(images=image, return_tensors="pt")
41
  outputs = model(**inputs)
42
+ last_hidden_states = outputs.last_hidden_state
 
 
43
  ```
44
 
45
  Currently, both the feature extractor and model support PyTorch. Tensorflow and JAX/FLAX are coming soon, and the API of ViTFeatureExtractor might change.