FLIP-base-32

CLIP model post-trained on 80M human face images.

Trained with TencentPretrain framework on 8 * A100 GPUs:

python3 pretrain.py --dataset_path faceclip.pt \
	--pretrained_model_path models/clip-b32.bin \
	--output_model_path models/faceclip-b32.bin \
	--config_path models/clip/base-32_config.json \
	--vocab_path vocab.json --merges_path merges.txt --tokenizer clip \
	--world_size 8 --gpu_ranks 0 1 2 3 4 5 6 7 --data_processor clip --accumulation_steps 8 --learning_rate 2e-5 \
	--total_steps 200000 --save_checkpoint_steps 20000 --batch_size 160 --report_steps 500

How to use:

from PIL import Image
import requests

from transformers import CLIPProcessor, CLIPModel

model = CLIPModel.from_pretrained("P01son/FaceCLIP-base-32")
processor = CLIPProcessor.from_pretrained("P01son/FaceCLIP-base-32")

url = "http://images.cocodataset.org/val2017/000000039769.jpg"
image = Image.open(requests.get(url, stream=True).raw)

inputs = processor(text=["a photo of a cat", "a photo of a dog"], images=image, return_tensors="pt", padding=True)

outputs = model(**inputs)
logits_per_image = outputs.logits_per_image # this is the image-text similarity score
probs = logits_per_image.softmax(dim=1) # we can take the softmax to get the label probabilities