spleeter-gpu is only using CPU to train. Am I doing something wrong?

I'm trying to train spleeter using my own stems but spleeter is only using my CPU

I have the following config:

Hardware
GTX1070
3600XT
16GB RAM

Software
Windows 10 x64 20H2 (19042.685)
Miniconda3-py37_4.9.2
spleeter-gpu 1.5.3
Nvidia driver 461.09
Cuda 10.0
Cudnn 7.4.1
I also included the folder C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\bin to PATH

I'm running spleeter using this command:

**set CUDA_VISIBLE_DEVICES=1 & spleeter train -p "C:/training/stems.json" -d "C:/training/stems" --verbose**

I'll attach my config files and the cmd output below. In this example, I'm only using one file to train and one to validation

[files.zip](https://github.com/deezer/spleeter/files/5822349/files.zip)

Start by inspecting files.zip and the supplied spleeter train command, then compare the reported Windows, NVIDIA driver, CUDA, and cuDNN versions with the spleeter-gpu 1.5.3 setup. Done means identifying why this training invocation uses the CPU and documenting a confirmed GPU-enabled configuration or the missing setup requirement.

Written by the indexing model from the issue text.