Views
No views yet
| Architecture | ViT-B-16 |
| Training data | CC12M (10.9M samples) |
| Epochs | 32 |
| Batch size | 512 x 8 GPUs |
| Vision pooling | Average |
| Text pooling | Average |
| SAM loss ratio | 0.1 |
| SAM regions topk | 10 |
| Softplus tau | 0.001 |
| Softplus alpha | 0.75 |
1import torch
2import open_clip
3
4model, _, preprocess = open_clip.create_model_and_transforms("ViT-B-16")
5
6# Load PowerCLIP checkpoint
7ckpt = torch.load("epoch_32.pt", map_location="cpu")
8model.load_state_dict(ckpt["state_dict"], strict=False)
9
10# Switch to average pooling (PowerCLIP default)
11model.visual.pool_type = "avg"
12model.text.pool_type = "avg"
13model.eval()