Spaces:
Running
on
Zero
Running
on
Zero
| import torch | |
| import torch.nn as nn | |
| from transformers import SiglipVisionModel, SiglipImageProcessor, SiglipVisionConfig | |
| class SiglipVisionTower(nn.Module): | |
| def __init__(self, vision_tower, args, delay_load=False): | |
| super().__init__() | |
| self.is_loaded = False | |
| self.vision_tower_name = vision_tower | |
| self.select_layer = args.mm_vision_select_layer | |
| self.select_feature = getattr(args, 'mm_vision_select_feature', 'patch') | |
| if not delay_load: | |
| self.load_model() | |
| elif getattr(args, 'unfreeze_mm_vision_tower', False): | |
| self.load_model() | |
| else: | |
| self.cfg_only = SiglipVisionConfig.from_pretrained(self.vision_tower_name) | |
| def load_model(self, device_map=None): | |
| if self.is_loaded: | |
| print('{} is already loaded, `load_model` called again, skipping.'.format(self.vision_tower_name)) | |
| return | |
| self.image_processor = SiglipImageProcessor.from_pretrained(self.vision_tower_name) | |
| self.vision_tower = SiglipVisionModel.from_pretrained(self.vision_tower_name, device_map=device_map) | |
| self.vision_tower.requires_grad_(False) | |
| self.is_loaded = True | |
| def feature_select(self, image_forward_outs): | |
| image_features = image_forward_outs.hidden_states[self.select_layer] | |
| if self.select_feature == 'patch': | |
| image_features = image_features[:, 1:] | |
| elif self.select_feature == 'cls_patch': | |
| image_features = image_features | |
| else: | |
| raise ValueError(f'Unexpected select feature: {self.select_feature}') | |
| return image_features | |
| def forward(self, images): | |
| if type(images) is list: | |
| image_features = [] | |
| for image in images: | |
| image_forward_out = self.vision_tower(image.to(device=self.device, dtype=self.dtype).unsqueeze(0), output_hidden_states=True) | |
| image_feature = self.feature_select(image_forward_out).to(image.dtype) | |
| image_features.append(image_feature) | |
| else: | |
| image_forward_outs = self.vision_tower(images.to(device=self.device, dtype=self.dtype), output_hidden_states=True) | |
| image_features = self.feature_select(image_forward_outs).to(images.dtype) | |
| return image_features | |
| def dummy_feature(self): | |
| return torch.zeros(1, self.hidden_size, device=self.device, dtype=self.dtype) | |
| def dtype(self): | |
| return self.vision_tower.dtype | |
| def device(self): | |
| return self.vision_tower.device | |
| def config(self): | |
| if self.is_loaded: | |
| return self.vision_tower.config | |
| else: | |
| return self.cfg_only | |
| def hidden_size(self): | |
| return self.config.hidden_size | |
| def num_patches_per_side(self): | |
| return self.config.image_size // self.config.patch_size | |
| def num_patches(self): | |
| return (self.config.image_size // self.config.patch_size) ** 2 | |
| class SiglipVisionTowerS2(SiglipVisionTower): | |
| def __init__(self, vision_tower, args, delay_load=False): | |
| super().__init__(vision_tower, args, delay_load) | |
| self.s2_scales = getattr(args, 's2_scales', '336,672,1008') | |
| self.s2_scales = list(map(int, self.s2_scales.split(','))) | |
| self.s2_scales.sort() | |
| self.s2_split_size = self.s2_scales[0] | |
| self.s2_image_size = self.s2_scales[-1] | |
| try: | |
| from s2wrapper import forward as multiscale_forward | |
| except ImportError: | |
| raise ImportError('Package s2wrapper not found! Please install by running: \npip install git+https://github.com/bfshi/scaling_on_scales.git') | |
| self.multiscale_forward = multiscale_forward | |
| # change resize/crop size in preprocessing to the largest image size in s2_scale | |
| if not delay_load or getattr(args, 'unfreeze_mm_vision_tower', False): | |
| self.image_processor.size['shortest_edge'] = self.s2_image_size | |
| self.image_processor.crop_size['height'] = self.image_processor.crop_size['width'] = self.s2_image_size | |
| def load_model(self, device_map=None): | |
| if self.is_loaded: | |
| print('{} is already loaded, `load_model` called again, skipping.'.format(self.vision_tower_name)) | |
| return | |
| self.image_processor = SiglipImageProcessor.from_pretrained(self.vision_tower_name) | |
| self.vision_tower = SiglipVisionModel.from_pretrained(self.vision_tower_name, device_map=device_map) | |
| self.vision_tower.requires_grad_(False) | |
| self.image_processor.size['shortest_edge'] = self.s2_image_size | |
| self.image_processor.crop_size['height'] = self.image_processor.crop_size['width'] = self.s2_image_size | |
| self.is_loaded = True | |
| def forward_feature(self, images): | |
| image_forward_outs = self.vision_tower(images.to(device=self.device, dtype=self.dtype), output_hidden_states=True) | |
| image_features = self.feature_select(image_forward_outs).to(images.dtype) | |
| return image_features | |
| def forward(self, images): | |
| if type(images) is list: | |
| image_features = [] | |
| for image in images: | |
| image_feature = self.multiscale_forward(self.forward_feature, image.unsqueeze(0), img_sizes=self.s2_scales, max_split_size=self.s2_split_size) | |
| image_features.append(image_feature) | |
| else: | |
| image_features = self.multiscale_forward(self.forward_feature, images, img_sizes=self.s2_scales, max_split_size=self.s2_split_size) | |
| return image_features | |
| def hidden_size(self): | |
| return self.config.hidden_size * len(self.s2_scales) | |