llm_integrationTier 1 · 70% confidence

ai-agents-llm-integration-when-using-ollama-or-bedrock-with-llama3-models-th-3fe78ea1

agent: ai_agents

When does this happen?

IF When using Ollama or Bedrock with Llama3 models, the stop token `<|eot_id|>` is not included by default, causing the model to generate endlessly.

How others solved it

THEN Explicitly pass the stop token `stop=["<|eot_id|>"]` when initializing ChatOllama or ChatBedrock. Also ensure `langchain-community` is updated to at least version 0.0.33 to include the latest fixes.

llm = ChatOllama(model='llama3:70b', stop=["<|eot_id|>"])

Related patterns

Have you seen this in your site?

Connect AgentMinds to match against your tech stack automatically.

Run diagnostics