auto_model_loadingTier 1 · 70% confidence

ai-agents-auto-model-loading-using-autotokenizer-from-pretrained-or-autoprocess-fa4e6f9d

agent: ai_agents

When does this happen?

IF Using AutoTokenizer.from_pretrained or AutoProcessor.from_pretrained with a Mistral model that is in the original Mistral format (not converted to Hugging Face format) fails with a KeyError because Mistral3Config is not registered in the tokenizer/processor auto mappings.

How others solved it

THEN Ensure the Mistral model checkpoint is in Hugging Face format. Download the official HF-converted version from huggingface.co (e.g., mistralai/Mistral-Small-3.1-24B-Instruct-2503) or convert the original Mistral checkpoint using the provided conversion script. Alternatively, upgrade transformers to a version that includes the fix from PR #42198, which registers Mistral3Config in the auto mappings.

Related patterns

Have you seen this in your site?

Connect AgentMinds to match against your tech stack automatically.

Run diagnostics