Implement model preloading in PoetryGenerationService for faster startup; remove redundant preload method from ModelManager

Remove unused import of PoetryGenerationService to clean up code

Refactor poem generation parameters; replace PoemRequest with GenerationParams for enhanced control and validation of generation settings

Refactor ModelManager and PoetryGenerationService for CPU optimization; enhance model initialization, improve text generation parameters, and implement explicit resource cleanup

Refactor PoetryGenerationService and ModelManager for CPU optimization; update model and tokenizer initialization, simplify prompt templates, and enhance text generation and cleaning processes

Refactor PoetryGenerationService to improve model preloading and poem generation; switch to max_new_tokens for better control and update model name for consistency

Add model preloading functionality to PoetryGenerationService for improved startup efficiency and error handling

Refactor PoetryGenerationService to enhance poem generation; switch to GPT2-medium model, improve prompt templates, and implement output cleaning for better formatting and quality

Refactor pipeline initialization in ModelManager to use device_map for automatic device selection

Update model name in PoetryGenerationService to use distilbert/distilgpt2 for improved compatibility in low-resource environments

Refactor PoetryGenerationService and ModelManager for improved resource efficiency; optimize model loading, implement caching, and enhance error handling

Optimize model loading and error handling in PoetryGenerationService; implement async poem generation and enhance application startup process

Improve model preloading in PoetryGenerationService with meaningful return value and enhanced error handling

Refactor PoetryGenerationService to streamline model initialization and improve error handling

Refactor model loading to use a consistent model name variable in PoetryGenerationService

Fix variable name for model loading in PoetryGenerationService

Add model preloading functionality to PoetryGenerationService and initialize on startup

Update model name in ModelManager to Llama-3.2-1B-Instruct

Remove optimized module usage from poetry generation service

Update model name in ModelManager to use Llama-1B-Instruct for improved performance

Enhance ModelManager and PoetryGenerationService with optimizations and new features

Fix initialization check in ModelManager to use class variable for singleton pattern

Refactor poetry generation to use singleton ModelManager for model and tokenizer initialization

Set tokenizer's pad token to end-of-sequence token for improved handling

Add initial project structure with FastAPI and poetry generation service