Reference only

rotary_embedding_torch is missing, or the rotary tensor dimensions do not match

A missing positional-encoding implementation is not a missing weight file. Once import works, inspect the API and tensor contract used by the actual model.

Scope and symptoms

A missing positional-encoding implementation is not a missing weight file. Once import works, inspect the API and tensor contract used by the actual model.

Search fragments; wording and context vary:

No module named 'rotary_embedding_torch'

Source-supported context

The rotary-embedding-torch project exposes its implementation through rotary_embedding_torch. Its operations belong to model computation; a similarly named file or an arbitrary dimension change is not evidence of compatibility. V04-ROTARY

Distinguish these cases

1. The declared distribution is missing from the active environment.

2. Import works but the node expects a different class, function or argument signature.

3. Execution reaches tensor operations with dimensions or position settings inconsistent with the model.

Suggested diagnostic sequence

This sequence is editorial guidance, not a diagnosis of your machine.

1. Establish the declared distribution, import name and version. Locate the failure at import, argument binding or tensor computation.

2. Follow the node's requirements for a missing package. For an API mismatch, check the maintainer's supported combination rather than installing an unrelated same-name repository.

3. Run the model author's smallest example and compare the input shape and positional arguments. Keep dimension values in diagnostic logs because they carry important evidence.

4. If the reference works and the workflow does not, compare upstream input and model configuration. Do not truncate tensors or alter weight dimensions simply to suppress the exception.

Completion checks

Import and the model's minimal positional operation succeed, and the workflow passes the same call. An output of an arbitrary shape is not sufficient.

Limits and cautions

No numerical-equivalence test for a particular model is claimed. Library updates, model configuration and quantization implementations require distinct checks.

Source review: 2026-09-22. No installation, GPU inference or user-environment repair was executed.

Original sources

Didn't solve it?

Check the next possible cause

The same symptom can come from a different cause. Work through these related entries in order.

  1. ModuleNotFoundError in ComfyUI even though the package is installedModuleNotFoundErrorA successful installation in a terminal does not show that ComfyUI uses the same Python. Separate the import name, distribution name and running interpreter before changing dependencies.
Paste your full log into the error finder

Was this page helpful?

Anonymous. Only a yes/no counter is stored; no account, IP address or device details.

Sources & references

Source-based editorial draft; no runtime verification. Version observations are scoped, and unknown wrapper identity is explicit.

01rotary-embedding-torch upstream repositorySource checked: 2026-09-22

Report an issue · cbb24b88-9053-513d-bc37-b61ab61820df