Does the Deep Learning Toolbox support automatic mixed-precision training?

Up to now the latest version of R2021a does not seem to support automatic mexed-precision training, such as NVIDIA open source apex support pytorch acceleration, future versions of matlab will correspond to support?

截至目前最新版本R2021a貌似没发现支持自动混合精度的训练，比如NVIDIA开源的apex支持pytorch加速，未来版本matlab会对应的支持吗？

[xingxingcui]

Still not supported in version R2021b！