redis.exceptions.ResponseError: GPU requested but Torch couldn't find CUDA

[Hanson0910]

Hi，When I use con.scriptset('imagenet_script', 'GPU', script) then raise redis.exceptions.ResponseError: GPU requested but Torch couldn't find CUDA.I had build the redisai with gpu.I can not fina any case with gpu in pyhton.

[hhsecond]

It's happening because an internal GPU check using torch is failing (a check similar to torch.cuda.is_available() from python). Could you provide the steps you have used to build RedisAI?

[Hanson0910]

Thanks for your prompt reply，this is my step:
1)git clone --recursive https://github.com/RedisAI/RedisAI;
2)bash get_deps.sh, it will download tf and torch..
3)ALL=1 make -C opt clean build
I build them in the python Virtual environment with python3.6 and torch1.5.1，the redis version=5.0.0

[rafie]

There's an error in the docs.
One should run the following to build RedisAI for GPU (that is, instead of steps 2, 3):

make -C opt fetch build GPU=1