# SCANVI: cannot reproduce predicted cell types as in the tutorial

**URL:** <https://discourse.scverse.org/t/scanvi-cannot-reproduce-predicted-cell-types-as-in-the-tutorial/635>\
**Category:** scvi-tools\
**Tags:** scanvi\
**Created:** [July 20, 2022, 7:54pm UTC](https://discourse.scverse.org/t/scanvi-cannot-reproduce-predicted-cell-types-as-in-the-tutorial/635 "2022-07-20T19:54:50Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![paupuigdevall](https://yyz1.discourse-cdn.com/flex035/user_avatar/discourse.scverse.org/paupuigdevall/32/323_2.png) [@paupuigdevall](https://discourse.scverse.org/u/paupuigdevall)\
**Post date:** [July 20, 2022, 7:54pm UTC](https://discourse.scverse.org/t/scanvi-cannot-reproduce-predicted-cell-types-as-in-the-tutorial/635/1 "2022-07-20T19:54:50Z")

</div>

Hi, I’m trying to reproduce the pipeline shown in here for the reference mapping with SCANVI:  
[tutorial](https://docs.scvi-tools.org/en/stable/tutorials/notebooks/scarches_scvi_tools.html#)

But for some reason I cannot reproduce the same predicted cell types, as the epsilon cell type is absent:

 ![Screenshot 2022-07-20 at 20.43.39](https://canada1.discourse-cdn.com/flex035/uploads/forum11/original/1X/e0378203539cae21c63e74137ff648244938b88e.png)

I’m using pip - scvi-tools==0.17.1 in Python 3.9.13

---

<div class="post-metadata">

**Author:** ![adamgayoso](https://yyz1.discourse-cdn.com/flex035/user_avatar/discourse.scverse.org/adamgayoso/32/100_2.png) [@adamgayoso](https://discourse.scverse.org/u/adamgayoso)\
**Post date:** [October 5, 2022, 11:10pm UTC](https://discourse.scverse.org/t/scanvi-cannot-reproduce-predicted-cell-types-as-in-the-tutorial/635/2 "2022-10-05T23:10:50Z")

</div>

There is some stochasticity, but we are looking into making scanvi more stable. Thanks for flagging.