3 个仓库
Layers that perform affine transformations via matrix multiplication and bias addition.
Distinct from Linear Mixing Layers: Distinct from Linear Mixing Layers by focusing on general size transformation (input to output) rather than mixing dimensions
Explore 3 awesome GitHub repositories matching artificial intelligence & ml · Linear Transformation Layers. Refine with filters or upvote what's useful.
Flashlight is a standalone C++ machine learning library and tensor library used for building and training neural networks. It functions as a comprehensive neural network framework and automatic differentiation engine, providing the tools to construct computation graphs and calculate gradients via backpropagation. The project serves as a distributed training framework, utilizing all-reduce operations to synchronize gradients and parameters across multiple compute nodes and devices. It distinguishes itself through deep integration of high-performance tensor manipulation, native device memory in
Implements linear transformation layers that use matrix multiplication and optional bias to transform input tensor sizes.
PerceptualSimilarity 是一个深度学习框架,旨在量化和评估图像之间的感知距离。它提供了一个系统,通过使用深度特征表示而不是像素级差异来衡量两张图像或图像块在人类视觉中看起来有多相似。 该项目实现了一个可微分距离度量,作为损失函数,允许通过反向传播优化图像像素以达到目标视觉外观。它包括一个可训练的线性层,可以应用于冻结的深度特征,以学习与人类感知一致的加权距离度量。 该框架涵盖了图像质量评估、相似度度量训练和计算机视觉基准测试方面的广泛功能。模型准确性通过使用诸如二选一强制选择测试等框架,将预测的距离分数与人类判断数据集进行比较来评估。
Uses a trainable linear layer on top of frozen features to learn weighted human-perceptual distances.
DeepSeek-VL 是一个多模态大型语言模型和图像到文本推理引擎。它作为一个视觉-语言模型和视觉问答系统,集成了视觉感知与语言推理,以理解和描述图像。 该项目支持多模态图像理解和文档图像分析,特别是处理网页截图和技术图表。它提供了视觉对话 AI 的功能,允许用户与视觉数据交互以提取见解,并跨不同类型的视觉信息执行复杂的推理。 该系统利用视觉-语言 Transformer 架构,将视觉 Transformer 与大型语言模型相结合。它采用多模态指令微调和投影层,将视觉特征向量与语言模型的嵌入空间对齐,以进行自回归文本生成。
Employs a learnable projection layer to align visual feature vectors with the language model's embedding space.