nielsr HF Staff commited on
Commit
4fada13
·
verified ·
1 Parent(s): 10f22aa

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +19 -1
README.md CHANGED
@@ -24,6 +24,24 @@ pip install -q git+https://github.com/huggingface/transformers.git
24
 
25
  Next you can use it like so:
26
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
27
  ```python
28
  import requests
29
  import torch
@@ -38,7 +56,7 @@ url = "http://images.cocodataset.org/val2017/000000039769.jpg"
38
  image = Image.open(requests.get(url, stream=True).raw)
39
  labels = ["a photo of a cat", "a photo of a dog", "a photo of a car"]
40
 
41
- inputs = processor(text=labels, images=image, return_tensors="pt", truncation=True, padding="max_length", max_length=77)
42
 
43
  outputs = model(**inputs)
44
  logits_per_image = outputs.logits_per_image
 
24
 
25
  Next you can use it like so:
26
 
27
+ ```python
28
+ import torch
29
+ from transformers import pipeline
30
+
31
+ clip = pipeline(
32
+ task="zero-shot-image-classification",
33
+ model="facebook/metaclip-2-worldwide-huge-quickgelu",
34
+ torch_dtype=torch.bfloat16,
35
+ device=0
36
+ )
37
+ labels = ["a photo of a cat", "a photo of a dog", "a photo of a car"]
38
+
39
+ results = clip("http://images.cocodataset.org/val2017/000000039769.jpg", candidate_labels=labels)
40
+ print(results)
41
+ ```
42
+
43
+ In case you want to perform pre- and postprocessing yourself, you can use the `AutoModel` API:
44
+
45
  ```python
46
  import requests
47
  import torch
 
56
  image = Image.open(requests.get(url, stream=True).raw)
57
  labels = ["a photo of a cat", "a photo of a dog", "a photo of a car"]
58
 
59
+ inputs = processor(text=labels, images=image, return_tensors="pt", padding=True)
60
 
61
  outputs = model(**inputs)
62
  logits_per_image = outputs.logits_per_image