XiaokunDing
Results
1
comments of
XiaokunDing
When exporting a model from PyTorch to ONNX using float16 precision, there is a significant difference in the output of the following operator. ` // Attention heads [sq, b, h]...