怎么得到shape=(b, l, d)的image feature

[PangziZhang523]

model, image_processor = create_model_from_pretrained(model_name='ViT-SO400M-14-SigLIP-384', pretrained=self.image_tower_name,
image_mean=([0.5,0.5,0.5]), image_std=([0.5,0.5,0.5]), image_interpolation="bicubic", image_resize_mode="squash")
image_features = model.encode_image(images) # [1, 1152]

这样得到的feature是 （1， 1152），序列长度怎么得到，用model.forward_features(images) 报错timm ttributeError: 'CustomTextCLIP' object has no attribute 'forward_features'

[PangziZhang523]

请教下diva的ViT-SO400M-14-SigLIP-384 的token长度是怎么做到121的呀