Spaces:
Running
on
Zero
Running
on
Zero
File size: 7,219 Bytes
e52682b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 |
import os
import torch
import torch.nn as nn
from transformers import (
CLIPVisionModel, CLIPImageProcessor, CLIPVisionConfig,
SiglipVisionModel, SiglipImageProcessor, SiglipVisionConfig
)
from .beats.BEATs import BEATsConfig, BEATs
class CLIPVisionTower(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()
else:
self.cfg_only = CLIPVisionConfig.from_pretrained(self.vision_tower_name)
def load_model(self):
self.image_processor = CLIPImageProcessor.from_pretrained(self.vision_tower_name)
self.vision_tower = CLIPVisionModel.from_pretrained(self.vision_tower_name)
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
@torch.no_grad()
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
@property
def dummy_feature(self):
return torch.zeros(1, self.hidden_size, device=self.device, dtype=self.dtype)
@property
def dtype(self):
return self.vision_tower.dtype
@property
def device(self):
return self.vision_tower.device
@property
def config(self):
if self.is_loaded:
return self.vision_tower.config
else:
return self.cfg_only
@property
def hidden_size(self):
return self.config.hidden_size
@property
def num_patches(self):
return (self.config.image_size // self.config.patch_size) ** 2
@property
def num_patches_per_side(self):
return self.config.image_size // self.config.patch_size
@property
def image_size(self):
return self.config.image_size
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()
else:
self.cfg_only = SiglipVisionConfig.from_pretrained(self.vision_tower_name)
def load_model(self):
self.image_processor = SiglipImageProcessor.from_pretrained(self.vision_tower_name)
self.vision_tower = SiglipVisionModel.from_pretrained(self.vision_tower_name)
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
else:
raise ValueError(f'Unexpected select feature: {self.select_feature}')
return image_features
@torch.no_grad()
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
@property
def dummy_feature(self):
return torch.zeros(1, self.hidden_size, device=self.device, dtype=self.dtype)
@property
def dtype(self):
return self.vision_tower.dtype
@property
def device(self):
return self.vision_tower.device
@property
def config(self):
if self.is_loaded:
return self.vision_tower.config
else:
return self.cfg_only
@property
def hidden_size(self):
return self.config.hidden_size
@property
def num_patches(self):
return (self.config.image_size // self.config.patch_size) ** 2
@property
def num_patches_per_side(self):
return self.config.image_size // self.config.patch_size
@property
def image_size(self):
return self.config.image_size
def build_vision_tower(vision_tower_cfg, **kwargs):
vision_tower = getattr(vision_tower_cfg, 'mm_vision_tower', getattr(vision_tower_cfg, 'vision_tower', None))
if 'clip' in vision_tower:
vision_tower = CLIPVisionTower(vision_tower, args=vision_tower_cfg, **kwargs)
elif 'siglip' in vision_tower:
vision_tower = SiglipVisionTower(vision_tower, args=vision_tower_cfg, **kwargs)
else:
raise ValueError(f'Unknown vision tower: {vision_tower}')
#print(vision_tower)
return vision_tower
def build_audio_tower(audio_tower_cfg, delay_load=False, **kwargs):
audio_tower = getattr(audio_tower_cfg, 'mm_audio_tower', getattr(audio_tower_cfg, 'audio_tower', None))
if not delay_load:
beats_checkpoint = torch.load(audio_tower, map_location='cpu')
if 'cfg' in beats_checkpoint:
beats_cfg = BEATsConfig(beats_checkpoint['cfg'])
else:
beats_cfg = BEATsConfig()
beats = BEATs(beats_cfg)
if not audio_tower.endswith('.bin'):
print(beats.load_state_dict(beats_checkpoint['model']))
else:
filtered_checkpoint = {}
prefix = 'model.audio_tower.'
for key, value in beats_checkpoint.items():
if key.startswith(prefix):
new_key = key[len(prefix):] # 去除前缀
filtered_checkpoint[new_key] = value
print(beats.load_state_dict(filtered_checkpoint, strict=False))
else:
beats_cfg = BEATsConfig()
beats = BEATs(beats_cfg)
return beats, beats_cfg
|