Riva speech nmt t2t issue

Please provide the following information when requesting support.

Hardware - GPU (Geforce 4090)
Hardware - CPU (AMD Ryzen 7 5700G with Radeon Graphics)
Operating System
TLT Version (if relevant)

Using Riva python NeuralMachineTranslationClientclient to do t2t translation, I encounter some situations in the t2t conversion.
see below image:

I try to translate Chinese characters like"看" convert to English。but I get “The first is the 100th”。 Any suggestions to avoid such kind issues?