CLIP-ViT-B/32 on the eight image classication tasks
Collection
if you find these models helpful, consider citing [our paper](https://arxiv.org/abs/2406.03280) • 9 items • Updated • 2
# Load model directly
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("tanganke/clip-vit-base-patch32_dtd")
model = AutoModel.from_pretrained("tanganke/clip-vit-base-patch32_dtd", device_map="auto")Quick Links
Model Card
Model Details
- Architecture: ViT-Base with patch size 32
- Training Data: DTD dataset
Training Details
Adam Optimizer with a constant learning rate 1e-5 for 4000 steps training (batch_size=32). Only the vision encoder is fine-tuned.
Evaluation Results
- pre-trained: 0.4388
- fine-tuned: 0.7973
Usage
load vision model
from transformers import CLIPVisionModel
vision_model = CLIPVisionModel.from_pretrained('tanganke/clip-vit-base-patch32_dtd')
substitute the vision encoder of clip
from transformers import CLIPModel
clip_model = CLIPModel.from_pretrained("openai/clip-vit-base-patch32")
clip_model.vision_model.load_state_dict(vision_model.vision_model.state_dict())
- Downloads last month
- 2,404
Model tree for tanganke/clip-vit-base-patch32_dtd
Base model
openai/clip-vit-base-patch32
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("feature-extraction", model="tanganke/clip-vit-base-patch32_dtd")