relay 本质上在TVM中就是一种模型描述的IR, 将各种模型编译成relay后再进行一步优化
示例用法: tvm.relay.frontend.from_onnx("/path/to/onnx/model") 加载onnx模型
tvm.relay.frontend
Functions:
|
| Convert from caffe model into compatible relay Function. |
|
| Load caffe2 graph which contains init_net and predict_net into Relay Function. |
|
| Convert from coreml model into Relay Function. |
|
| Convert from Darknet’s model into compatible relay Function. |
|
| Convert keras model to relay Function. |
|
| Convert from MXNet”s model into compatible relay Function. |
|
| Convert a ONNX model into an equivalent Relay Function. |
|
| Load PyTorch model in the form of a scripted PyTorch model and convert into relay. |
|
| Load tensorflow graph which is a python tensorflow graph object into relay. |
|
| Convert from tflite model into compatible relay Function. |
|
| Quantized conv2d bias |
博客介绍了在TVM中,relay是一种模型描述的IR,可将各种模型编译成relay后进行优化,还给出了使用tvm.relay.frontend.from_onnx加载onnx模型的示例用法。

2653

被折叠的 条评论
为什么被折叠?



