ferraus accelerate install
Install GPU execution-provider libraries for ONNX Runtime
Synopsis
ferraus accelerate install [options]
What it does
Downloads the ONNX Runtime GPU execution-provider libraries that match this build into the user cache and verifies them. On macOS CoreML is built in and on Windows DirectML is built in, so nothing is downloaded for them. The CUDA provider requires NVIDIA's CUDA and cuDNN runtimes, which are covered by the NVIDIA EULA; pass --accept-eula (or -y) to confirm you accept it.
Examples
Fetch the CUDA provider on Linux or Windows
ferraus accelerate install --accept-eula
Install the CUDA provider; -y also accepts the NVIDIA EULA
ferraus accelerate install --provider cuda -y
On macOS: report that CoreML is built in (nothing to download)
ferraus accelerate install --provider coreml --json
Options
| Flag | Type | Default | Meaning |
|---|---|---|---|
--provider | auto | cuda | coreml | directml | auto | Which provider to install (auto: CUDA on Linux/Windows, report CoreML on macOS) |
--accept-eula | flag | — | Accept the NVIDIA EULA for the CUDA runtime libraries |
Global options
Every command also accepts the global options (-o, --json, --dry-run, --overwrite, --progress, ...). See Global options.
Exit codes
| Code | Meaning |
|---|---|
| 0 | Success |
| 1 | Processing failed |
| 2 | Usage or parameter error |
| 3 | Input missing or unreadable |
| 5 | Model not cached and downloads disabled |
| 6 | Model download or verification failed |
| 8 | Cancelled |
| 10 | Network error |
| 70 | Internal error (bug) |
Common mistakes
- Expecting human-readable text when scripting: add
--jsonfor a stable machine result.
For models
- MCP tool:
ferraus_accelerate_install - Schema:
ferraus schema accelerate install - Pipeline:
{ op: accelerate install, args: { ... } }(flag names with_instead of-) - fx stage: no
{"command": "accelerate install", "defaults": {"provider": "auto"}}