Heterogeneous Deep Learning Compiler Framework!
The MERA software framework from EdgeCortix allows for the seamless compilation and execution of standard or custom CNNs (Convolutional Neural Networks) developed with industry-standard frameworks on heterogeneous platforms, including the EdgeCortix SAKURA AI coprocessor, by installing it from the public pip repository. MERA is equipped with Apache TVM and provides a simple API to seamlessly perform graph compilation and inference of deep neural networks using SAKURA's DNA AI engine. Additionally, it offers profiling tools, code generators, and runtime, enabling the deployment of pre-trained deep neural networks after a simple calibration and quantization step. MERA supports models that are directly quantized in deep learning frameworks such as Pytorch and Tensorflowlite.
Inquire About This Product
basic information
Diverse Operator Support - Standard convolution and Depthwise convolution - Stride and dilation - Symmetric/Asymmetric padding - Max pooling, average pooling - ReLU, ReLU6, LeakyReLU, H-Swish, and H-Sigmoid - Upsampling and downsampling - Residual connections, branches, etc. Drop-in Replacement for GPU - Interfaces for Python and C++ - Native support for PyTorch and TensorFlow-lite - No retraining required - Supports high-resolution input INT8 Bit Quantization - Post-training calibration and quantization - Maintains high accuracy Built-in Simulator - Allows inference simulation in x86 environments without using SAKURA devices - Estimates inference latency and throughput under various conditions For more details, please contact us through our website. Reference link: https://www.edgecortix.com/
Price range
Delivery Time
Applications/Examples of results
For more details, please download the PDF or contact us.
catalog（1）
Download All Catalogs
Company information
We offer products such as the deep learning inference IP core "Dynamic Neural Accelerator (DNA)" characterized by high computational power, ultra-low latency, and a scalable inference engine, as well as the edge AI coprocessor SoC "EdgeCortix SAKURA," which achieves industry-leading computational efficiency and latency, and the software framework "EdgeCortix MERA," which supports various platforms. Please feel free to contact us if you have any inquiries.