Skip to content

Could not start backend on Jetson AGX Orin #466

New issue

Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? # to your account

Closed
2 of 4 tasks
HiddenPeak opened this issue Jan 4, 2025 · 3 comments · May be fixed by #467
Closed
2 of 4 tasks

Could not start backend on Jetson AGX Orin #466

HiddenPeak opened this issue Jan 4, 2025 · 3 comments · May be fixed by #467

Comments

@HiddenPeak
Copy link

System Info

Jetson Orin AGX 64G
SM 8.7
cargo 1.75.0 (1d8b05cdd 2023-11-20)

Information

  • Docker
  • The CLI directly

Tasks

  • An officially supported command
  • My own modifications

Reproduction

  1. curl https://sh.rustup.rs -sSf | sh
  2. sudo apt-get install libssl-dev gcc -y
  3. git clone https://github.com/huggingface/text-embeddings-inference.git
  4. cd text-embeddings-inference
  5. cargo install --path router -F candle-cuda -F http --no-default-features
  6. model=BAAI/bge-reranker-large
  7. text-embeddings-router --model-id $model --port 8080

Expected behavior

Could not start Candle backend: Could not start backend: Runtime compute cap 87 is not compatible with compile time compute cap 87
Error: Could not create backend

Caused by:
    Could not start backend: Could not start a suitable backend
@HiddenPeak
Copy link
Author

in backends/candle/src/compute_cap.rs

I try to add SM87 support, and I need some time to test it.

If anyone want to try it,use [this] (https://github.com/HiddenPeak/text-embeddings-inference ) befor it merged.

@HiddenPeak
Copy link
Author

74f61043ee2b7a9da75e93547f397b7
6519f72d3f88800fd8e89705b2d5fe1
fc2f46532bbe2545f4f7f2c55388190
It works

@HiddenPeak
Copy link
Author

on Jetpack 6.1 with cuda 12.6

# for free to join this conversation on GitHub. Already have an account? # to comment
Labels
None yet
Projects
None yet
Development

Successfully merging a pull request may close this issue.

1 participant