Dimension is not supported by cohere/Hugging Face/...

[MFHLLLL]

R2RException: {'message': "Error getting embeddings: litellm.UnsupportedParamsError: Setting {'dimensions': 1024} is not supported by cohere. To drop it from the call, set litellm.drop_params = True.", 'error_type': 'R2RException'}

[NolanTrem]

You should set dimension equal to 'nan' in your config. I'll try to make this more clear in the docs, anywhere that you would suggest it be added?

See the LM Studio implementation section here for an example:
https://r2r-docs.sciphi.ai/self-hosting/local-rag#local-configuration-details-1

[xXMrNidaXx]

Dimension mismatch errors with Cohere/HuggingFace embeddings — here's the fix:

The problem:
Different embedding models output different dimensions:

• OpenAI: 1536 (ada-002) or 3072 (text-embedding-3-large)
• Cohere: 1024 or 4096
• HuggingFace: varies by model

Solution 1: Match dimensions in config

# r2r.toml
[embedding]
model = "cohere/embed-english-v3.0"
dimension = 1024  # Must match model output!

Solution 2: Recreate vector collection

# If you changed models, recreate the collection
r2r db reset
r2r ingest --files your_docs/

Solution 3: Use model with matching dims

# Check your model's output dimension
from sentence_transformers import SentenceTransformer
model = SentenceTransformer("your-model")
print(model.get_sentence_embedding_dimension())

Common dimension mismatches:

Model	Dimension
OpenAI ada-002	1536
Cohere v3	1024
all-MiniLM	384
BGE-large	1024

We handle multi-model setups at RevolutionAI. What model are you trying to use?