diff --git a/Dockerfile.train b/Dockerfile.train
index 3fbf3dcb..dba45e6e 100644
--- a/Dockerfile.train
+++ b/Dockerfile.train
@@ -2,7 +2,7 @@
 # You can train "acoustic models" with audio + Tensorflow, and
 # you can create "scorers" with text + KenLM.
 
-FROM ubuntu:20.04 AS kenlm-build
+FROM nvcr.io/nvidia/tensorflow:20.06-tf1-py3 AS kenlm-build
 ENV DEBIAN_FRONTEND=noninteractive
 
 RUN apt-get update && \