MMCV installation and missing-operator errors: check the build combination

The final exit code is not a diagnosis

MMCV documents full and lite distributions, operator support and environment-matched prebuilt packages. Without a matching wheel, installation may take a source-build path.1

python -m pip show mmcv mmcv-lite torch

Our suggested comparison includes the node's required MMCV major version, Python, Torch build and platform. A missing _ext is not automatically a missing Python installation. Check whether the selected distribution contains the required operators and whether native loading succeeded.

For a build failure, retain the concrete compiler error rather than only exit code 1. Choose an upstream-supported build combination or document the requirements for building in an isolated environment. Do not substitute a similarly named wheel or remove every Torch-related package as the first action.

Validate the trusted package, the particular operator required by the node, and a minimal workflow separately. Printing a version number only verifies the first of those checkpoints.

Related checks

• SAM and GroundingDINO failures: separate code, checkpoints and extensions
• Nunchaku still fails after installation: align the node, wheel and model

Research draft. No local/GPU reproduction or independent editorial sign-off. Upstream statements and our recommendations are distinguished.