Total runs:
1.8M
Run Growth:
382.5K
Growth Rate:
21.05%
Salesforce / blip2-flan-t5-xxl
Model's Last Updated: November 22 2024
image-text-to-text
Introduction of blip2-flan-t5-xxl
Model Details of blip2-flan-t5-xxl
BLIP-2, Flan T5-xxl, pre-trained only
BLIP-2 model, leveraging Flan T5-xxl (a large language model). It was introduced in the paper BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models by Li et al. and first released in this repository .
Disclaimer: The team releasing BLIP-2 did not write a model card for this model so this model card has been written by the Hugging Face team.
Model description
BLIP-2 consists of 3 models: a CLIP-like image encoder, a Querying Transformer (Q-Former) and a large language model.
The authors initialize the weights of the image encoder and large language model from pre-trained checkpoints and keep them frozen while training the Querying Transformer, which is a BERT-like Transformer encoder that maps a set of "query tokens" to query embeddings, which bridge the gap between the embedding space of the image encoder and the large language model.
The goal for the model is simply to predict the next text token, giving the query embeddings and the previous text.
This allows the model to be used for tasks like:
- image captioning
- visual question answering (VQA)
- chat-like conversations by feeding the image and the previous conversation as prompt to the model
Direct Use and Downstream Use
You can use the raw model for conditional text generation given an image and optional text. See the model hub to look for fine-tuned versions on a task that interests you.
Bias, Risks, Limitations, and Ethical Considerations
BLIP2-FlanT5 uses off-the-shelf Flan-T5 as the language model. It inherits the same risks and limitations from Flan-T5 :
Language models, including Flan-T5, can potentially be used for language generation in a harmful way, according to Rae et al. (2021). Flan-T5 should not be used directly in any application, without a prior assessment of safety and fairness concerns specific to the application.
BLIP2 is fine-tuned on image-text datasets (e.g. LAION ) collected from the internet. As a result the model itself is potentially vulnerable to generating equivalently inappropriate content or replicating inherent biases in the underlying data.
BLIP2 has not been tested in real world applications. It should not be directly deployed in any applications. Researchers should first carefully assess the safety and fairness of the model in relation to the specific context they’re being deployed within.
How to use
For code examples, we refer to the documentation , or refer to the snippets below depending on your usecase:
Running the model on CPU
Click to expand
import requests
from PIL import Image
from transformers import BlipProcessor, Blip2ForConditionalGeneration
processor = BlipProcessor.from_pretrained("Salesforce/blip2-flan-t5-xxl")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-flan-t5-xxl")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
Running the model on GPU
In full precision
Click to expand
# pip install accelerate
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-flan-t5-xxl")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-flan-t5-xxl", device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
In half precision (
float16
)
Click to expand
# pip install accelerate
import torch
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-flan-t5-xxl")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-flan-t5-xxl", torch_dtype=torch.float16, device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
In 8-bit precision (
int8
)
Click to expand
# pip install accelerate bitsandbytes
import torch
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-flan-t5-xxl")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-flan-t5-xxl", load_in_8bit=True, device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
Runs of Salesforce blip2-flan-t5-xxl on huggingface.co
5.0K
Total runs
0
24-hour runs
56
3-day runs
18
7-day runs
-3.2K
30-day runs
More Information About blip2-flan-t5-xxl huggingface.co Model
Url of blip2-flan-t5-xxl
blip2-flan-t5-xxl huggingface.co Url
Provider of blip2-flan-t5-xxl huggingface.co
Other API from Salesforce
Updated: August 01 2023
Total runs:
1.1M
Run Growth:
-446.1K
Growth Rate:
-40.66%
Total runs:
630.3K
Run Growth:
516.2K
Growth Rate:
98.28%
Total runs:
299.2K
Run Growth:
-898
Growth Rate:
-0.31%
Total runs:
286.7K
Run Growth:
47.2K
Growth Rate:
16.81%
Total runs:
251.1K
Run Growth:
8.2K
Growth Rate:
3.29%
Total runs:
186.8K
Run Growth:
37.5K
Growth Rate:
93.70%
Total runs:
181.1K
Run Growth:
57.5K
Growth Rate:
31.77%
Total runs:
169.1K
Run Growth:
-141.5K
Growth Rate:
-83.04%
Total runs:
96.6K
Run Growth:
-17.0K
Growth Rate:
-17.84%
Total runs:
83.6K
Run Growth:
32.8K
Growth Rate:
39.73%
Total runs:
64.7K
Run Growth:
15.6K
Growth Rate:
24.17%
Total runs:
58.3K
Run Growth:
3.6K
Growth Rate:
5.96%
Total runs:
54.3K
Run Growth:
18.7K
Growth Rate:
34.64%
Total runs:
41.7K
Run Growth:
-9.1K
Growth Rate:
-21.92%
Total runs:
35.7K
Run Growth:
-9.1K
Growth Rate:
-25.52%
Total runs:
35.2K
Run Growth:
-4.2K
Growth Rate:
-12.18%
Total runs:
33.0K
Run Growth:
11.9K
Growth Rate:
88.03%
Total runs:
31.0K
Run Growth:
8.0K
Growth Rate:
26.13%
Total runs:
30.5K
Run Growth:
5.2K
Growth Rate:
16.96%
Total runs:
26.3K
Run Growth:
3.7K
Growth Rate:
14.24%
Total runs:
22.7K
Run Growth:
-7.0K
Growth Rate:
-30.68%
Total runs:
22.0K
Run Growth:
9.6K
Growth Rate:
44.62%
Total runs:
18.7K
Run Growth:
12.5K
Growth Rate:
66.73%
Total runs:
12.5K
Run Growth:
-4.4K
Growth Rate:
-35.61%
Total runs:
9.3K
Run Growth:
4.0K
Growth Rate:
44.98%
Total runs:
8.3K
Run Growth:
-8.2K
Growth Rate:
-98.87%
Total runs:
8.3K
Run Growth:
-571
Growth Rate:
-7.48%
Total runs:
8.0K
Run Growth:
4.5K
Growth Rate:
56.20%
Total runs:
7.0K
Run Growth:
-11.5K
Growth Rate:
-163.50%
Total runs:
5.8K
Run Growth:
-2.1K
Growth Rate:
-39.01%
Total runs:
3.9K
Run Growth:
-499
Growth Rate:
-12.97%
Total runs:
3.9K
Run Growth:
569
Growth Rate:
14.81%
Total runs:
3.4K
Run Growth:
-427
Growth Rate:
-13.11%
Total runs:
3.1K
Run Growth:
2.7K
Growth Rate:
90.33%
Total runs:
3.1K
Run Growth:
1.0K
Growth Rate:
32.82%
Total runs:
3.0K
Run Growth:
1.9K
Growth Rate:
61.78%
Total runs:
2.9K
Run Growth:
1.2K
Growth Rate:
45.56%
Total runs:
2.2K
Run Growth:
-2.2K
Growth Rate:
-90.01%
Total runs:
2.1K
Run Growth:
1.3K
Growth Rate:
59.40%
Total runs:
2.0K
Run Growth:
0
Growth Rate:
0.00%
Total runs:
1.9K
Run Growth:
1.3K
Growth Rate:
79.28%
Total runs:
1.6K
Run Growth:
1.1K
Growth Rate:
76.96%
Total runs:
1.5K
Run Growth:
1.4K
Growth Rate:
95.54%
Total runs:
1.5K
Run Growth:
1.1K
Growth Rate:
72.96%
Total runs:
1.5K
Run Growth:
-6
Growth Rate:
-0.43%
Total runs:
1.5K
Run Growth:
-3.5K
Growth Rate:
-258.73%
Total runs:
1.4K
Run Growth:
0
Growth Rate:
0.00%
Total runs:
1.4K
Run Growth:
-1.2K
Growth Rate:
-86.65%
Total runs:
1.3K
Run Growth:
3
Growth Rate:
0.25%
Total runs:
1.2K
Run Growth:
-1.4K
Growth Rate:
-116.67%
Total runs:
1.1K
Run Growth:
-518
Growth Rate:
-43.49%
Total runs:
1.1K
Run Growth:
351
Growth Rate:
33.21%
Total runs:
1.1K
Run Growth:
-171
Growth Rate:
-18.35%
Total runs:
1.0K
Run Growth:
-49
Growth Rate:
-4.64%
Total runs:
1000
Run Growth:
47
Growth Rate:
5.10%
Total runs:
839
Run Growth:
267
Growth Rate:
32.29%
Total runs:
833
Run Growth:
-133
Growth Rate:
-16.18%
Total runs:
824
Run Growth:
60
Growth Rate:
7.22%
Total runs:
710
Run Growth:
323
Growth Rate:
47.22%
Total runs:
538
Run Growth:
0
Growth Rate:
0.00%
Total runs:
469
Run Growth:
-201
Growth Rate:
-42.49%
Total runs:
423
Run Growth:
43
Growth Rate:
10.17%
Total runs:
406
Run Growth:
105
Growth Rate:
30.09%
Total runs:
233
Run Growth:
-474
Growth Rate:
-234.65%
Total runs:
222
Run Growth:
-204
Growth Rate:
-92.73%
Total runs:
199
Run Growth:
-57
Growth Rate:
-28.93%
Total runs:
185
Run Growth:
47
Growth Rate:
25.41%
Total runs:
183
Run Growth:
-9.1K
Growth Rate:
-5252.02%
Total runs:
167
Run Growth:
-585
Growth Rate:
-325.00%
Total runs:
157
Run Growth:
-400
Growth Rate:
-298.51%
Total runs:
134
Run Growth:
-305
Growth Rate:
-193.04%
Total runs:
119
Run Growth:
-8.9K
Growth Rate:
-7473.11%
Total runs:
117
Run Growth:
-377
Growth Rate:
-330.70%
Total runs:
116
Run Growth:
86
Growth Rate:
74.14%
Total runs:
113
Run Growth:
106
Growth Rate:
93.81%
Total runs:
108
Run Growth:
0
Growth Rate:
0.00%
Total runs:
107
Run Growth:
91
Growth Rate:
85.05%
Total runs:
105
Run Growth:
93
Growth Rate:
86.92%
Total runs:
100
Run Growth:
-326
Growth Rate:
-291.07%
Total runs:
97
Run Growth:
-137
Growth Rate:
-135.64%
Total runs:
96
Run Growth:
-143
Growth Rate:
-153.76%
Total runs:
87
Run Growth:
-100
Growth Rate:
-114.94%
Total runs:
85
Run Growth:
-202
Growth Rate:
-226.97%
Total runs:
73
Run Growth:
5
Growth Rate:
6.85%
Total runs:
56
Run Growth:
28
Growth Rate:
50.00%
Total runs:
53
Run Growth:
-85
Growth Rate:
-134.92%
Total runs:
40
Run Growth:
34
Growth Rate:
85.00%
Total runs:
39
Run Growth:
33
Growth Rate:
84.62%