CLIP算法的Loss详解 和 交叉熵CrossEntropy实现

CLIP算法的Loss详解 和 交叉熵CrossEntropy实现

CLIP：Contrastive Language–Image Pre-training(可对比语言-图像预训练算法)是OpenAI提出的多模态预训练的算法，在各种各样的**样本对(图像、文本)**上训练的神经网络。

具体参考：CLIP、OpenCLIP

其中，流程：

loss_i和loss_t的具体源码如下，参考 model.py：

    def forward(self, image, text):image_features = de_image(image)text_features = de_text(text)# normalized featuresimage_features = image_features / (dim=1, keepdim=True)text_features = text_features / (dim=1, keepdim=True)# cosine similarity as logitslogit_scale = self.p()logits_per_image = logit_scale * image_features @ text_features.t()logits_per_text = logits_per_image.t()# shape = [global_batch_size, global_batch_size]return logits_per_image, logits_per_text

其中，labels是torch.arange(batch_size, device=device).long()，参考train.py，具体如下

        _grad():for i, batch in enumerate(dataloader):images, texts = batchimages = (device=device, non_blocking=True)texts = (device=device, non_blocking=True)with autocast():image_features, text_features, logit_scale = model(images, texts)# features are accumulated in CPU tensors, otherwise GPU memory exhausted quickly# however, system RAM is easily exceeded and compute time becomes problematicall_image_features.append(image_features.cpu())all_text_features.append(text_features.cpu())logit_scale = an()logits_per_image = logit_scale * image_features @ text_features.t()logits_per_text = logits_per_image.t()batch_size = images.shape[0]labels = torch.arange(batch_size, device=device).long()total_loss = (F.cross_entropy(logits_per_image, labels) &#ss_entropy(logits_per_text, labels)) / 2

交叉熵函数：y就是label，x_softmax[i][y[i]]，表示在x_softmax中筛选第i个sample的第y[i]个值，作为log的输入，全部log负向求和，再求均值。

• y所对应的就是CLIP的np.arange(n)，也就是依次是第0个位置~第n-1个位置，计算log。

# 定义softmax函数
def softmax(x):p(x) / np.p(x))# 利用numpy计算
def cross_entropy_np(x, y):x_softmax = [softmax(x[i]) for i in range(len(x))]x_log = [np.log(x_softmax[i][y[i]]) for i in range(len(y))]loss = - np.sum(x_log) / len(y)return loss# 测试逻辑
x = [[1.9269, 1.4873, 0.9007, -2.1055]]
y = [[2]]
v1 = cross_entropy_np(x, y)
print(f"v1: {v1}")x = torch.unsqueeze(torch.Tensor(x), dim=0)
x = x.transpose(1, 2)  # CrossEntropy输入期望: Class放在第2维，Batch放在第1维y = torch.Tensor(y)
y = y.to(torch.long)  # label的类型为longv2 = F.cross_entropy(x, y, reduction="none")
print(f"v2: {v2}")

输出：

v1: 1.729491540989093
v2: tensor([[1.7295]])

参考：

• arxiv文章下载很慢怎么办？
• CLIP-对比图文多模态预训练的读后感
• CrossEntropy的numpy实现和Pytorch调用