# Optimized Generation Configurations for Different Use Cases CONFIGS = { "creative": { "temperature": 0.9, "top_k": 50, "top_p": 0.95, "repetition_penalty": 1.1, "description": "More creative and diverse outputs" }, "balanced": { "temperature": 0.7, "top_k": 40, "top_p": 0.9, "repetition_penalty": 1.3, "description": "Balanced creativity and coherence (recommended)" }, "focused": { "temperature": 0.5, "top_k": 30, "top_p": 0.85, "repetition_penalty": 1.5, "description": "More focused and deterministic" }, "factual": { "temperature": 0.3, "top_k": 20, "top_p": 0.8, "repetition_penalty": 1.4, "description": "Best for encyclopedia-style content" } } # Better prompts for small models PROMPT_TEMPLATES = { "article": "Wikipedia article about {topic}:\n\n", "definition": "{term} is defined as", "explanation": "Here is an explanation of {topic}:\n\n", "continuation": "{text}" }