TVM性能评估分析(三)

30
五月
2021

TVM性能评估分析(三)
在这里插入图片描述

Figure 1. TVM’s WebGPU backend close to native GPU performance when deploying models to the web.
在这里插入图片描述

Figure 2. WebGPU is to write shaders for primitive operators in deep neural networks
在这里插入图片描述

Figure 3. Build a WebGPU runtime inside TVM’s JS runtime
在这里插入图片描述

Figure 4. Comparing the execution of a full computational graph via TVM’s WebGPU backend and native targets
在这里插入图片描述

Figure 5. 2D convolution with data layout in NCHW4c and weight layout in OIHW4o4i. Left: The input tensor in NCHW4c layout. One moving filter of the kernel is colored in blue. One element of the input and kernel is colored in grey. Mid: The packed input and kernel in the grey block. Right: The output in NCHW4c layout. Inside the one element depicted, there are four packed elements in channel sub-dimension.
在这里插入图片描述

Figure 6. Workflow of running quantized models
在这里插入图片描述

Figure 7. A full deep learning compiler stack to support machine learning workloads for diverse hardware backends.
在这里插入图片描述

Figure 8. Golang Interface over TVM Runtime
在这里插入图片描述

Figure 9. Import, Compile, Integrate and Deploy

TAG

网友评论

共有访客发表了评论
请登录后再发布评论,和谐社会,请文明发言,谢谢合作! 立即登录 注册会员