Failed to deploy in NVIDIA H20

[DaiJianghai]

System Info

setup

docker image: ghcr.io/huggingface/text-embeddings-inference:hopper-1.7
gpu: H20
cuda version in host: 12.8

try

i try use command docker exec -it tei-bge-m3 nvidia-smi and found the gpu info is detected,
but when i use docker logs tei-bge-m3 then found some init error like `can't detect cuda device,... use cpu instead'

I want to know what caused such error and how can i deploy the embedding model(bge-m3) in GPU H20

Thanks!

Information

• Docker
• The CLI directly

Tasks

• An officially supported command
• My own modifications

Reproduction

docker pull ghcr.io/huggingface/text-embeddings-inference:hopper-1.7
use docker compose to serve model bge-m3

Expected behavior

I want to know what caused such error and how can i deploy the embedding model(bge-m3) in GPU H20