intfloat

/

multilingual-e5-large

like 1.07k

When I deploying E5 using TEI image " ghcr.io/huggingface/text-embeddings-inference:latest" with 1 pc of H200, it shows the error as below, telling that the compute cap 90 is not compatible with compute cap 80. But from the github documentation, it stated that the TEI image is compatible with compute cap 90.

{"timestamp":"2025-09-19T08:23:10.566970Z","level":"ERROR","message":"Could not start Candle backend: Could not start backend: Runtime compute cap 90 is not compatible with compile time compute cap 80","target":"text_embeddings_backend","filename":"backends/src/lib.rs","line_number":448}
Error: Could not create backend

Caused by:
Could not start backend: Could not start a suitable backend