TPM count matrix for Azimuth annotation

[seongguheo]

Dear Authors,

I am analyzing Smart-seq2 single cell RNA-seq data, which is a full length RNA-seq data.
TPM removes transcript length difference effects from the data.
So, I wonder if I can use the TPM count matrix for Azimuth annotation or I have to use the raw count matrix data since Azimuth performs normalization by itself.

And I wonder If there is a way around of Azimuth's normalization to avoid duplicating normalization in the case I use TPM.

Thank you so much for your help.
Luke