fireworks ai how config fireworks.json for llama-3

Configuration steps

Base model

Set the base_model to the appropriate Llama-3 model:

jsonCopy
{
  "base_model": "accounts/fireworks/models/llama-v3-8b"
}

or

jsonCopy
{
  "base_model": "accounts/fireworks/models/llama-v3-70b"
}

This specifies the Llama-3 model to be used 2.

Conversation configuration

Enable the conversation API by adding conversation_config:

jsonCopy
{
  "base_model": "accounts/fireworks/models/llama-v3-8b",
  "conversation_config": {
    "style": "llama-chat",
    "args": {
      "system_prompt": "You are a helpful assistant."
    }
  }
}

This configures the model for chat-based interactions 5.

JSON mode

To enforce structured outputs, use JSON mode:

jsonCopy
{
  "base_model": "accounts/fireworks/models/llama-v3-8b",
  "conversation_config": {
    "style": "llama-chat",
    "args": {
      "system_prompt": "You are a helpful assistant."
    }
  },
  "response_format": {
    "type": "json",
    "schema": {
      "type": "object",
      "properties": {
        "title": {"type": "string"},
        "description": {"type": "string"},
        "steps": {"type": "array"}
      }
    }
  }
}

This ensures the model's output adheres to the specified JSON schema 11.

Example configuration

Combining all elements, a complete fireworks.json for Llama-3 might look like:

jsonCopy
{
  "base_model": "accounts/fireworks/models/llama-v3-8b",
  "conversation_config": {
    "style": "llama-chat",
    "args": {
      "system_prompt": "You are a helpful assistant."
    }
  },
  "response_format": {
    "type": "json",
    "schema": {
      "type": "object",
      "properties": {
        "title": {"type": "string"},
        "description": {"type": "string"},
        "steps": {"type": "array"}
      }
    }
  }
}

This configuration sets up Llama-3 for chat interactions with structured JSON output.

Conclusion

Configure fireworks.json for Llama-3 by setting the base_model, enabling conversation_config, and specifying JSON mode for structured outputs.