.cursorrules

{
  "rules": {
    "context_initialization": {
      "description": "Starting point for each interaction, ensuring thorough understanding of project requirements and references",
      "steps": [
        "ALWAYS read `.notes/project_overview.md` and `.notes/task_list.md` before beginning any coding or documentation task",
        "ALWAYS confirm that you are operating under the guidelines specified for deep learning, diffusion models, and LLM development"
      ]
    },
    "operational_protocol": {
      "description": "How to approach tasks and maintain code quality",
      "before_action": [
        "Create a MECE (Mutually Exclusive, Collectively Exhaustive) task breakdown for clarity and efficient problem solving"
      ],
      "code_changes": [
        "Read relevant code sections before editing to preserve existing functionality",
        "Maintain type safety and best practices",
        "Ensure code is written according to PEP 8 style guidelines",
        "Implement GPU/mixed-precision optimizations when possible"
      ]
    },
    "ai_capabilities": {
      "description": "Expert AI skill set relevant to project",
      "capabilities": [
        "Extensive knowledge of deep learning with PyTorch",
        "Expertise in Transformers library for LLM and advanced token handling",
        "Skilled in Diffusers library for Stable Diffusion models",
        "Proficient in LoRA (Low-Rank Adaptation) fine-tuning techniques",
        "Adept at mixed precision training (torch.cuda.amp)",
        "Proficient at implementing attention mechanisms, positional encodings, and advanced text/image pipelines",
        "Ability to create object-oriented model architectures and functional data pipelines"
      ]
    },
    "project_technologies": {
      "description": "Technologies and libraries used in this project",
      "tech_stack": [
        "PyTorch",
        "Transformers",
        "Diffusers",
        "Accelerate",
        "LoRA fine-tuning",
        "StabilityAI Stable Diffusion XL Base 1.0",
        "StabilityAI Stable Diffusion XL Refiner 1.0",
        "Numpy",
        "TQDM",
        "TensorBoard or Weights & Biases for logging",
        "Python 3.9+"
      ]
    },
    "task_integration_and_debugging": {
      "description": "Ensure all changes are documented in directory structure, task list, and debug logs",
      "directives": [
        "After making any change, update the directory structure if necessary and reflect these changes in `.notes/task_list.md`",
        "Insert short, concise print statements for debugging at critical steps in scripts (e.g., 'Loading config...', 'Starting training...', 'Saving checkpoint...')",
        "Document every AI interaction thoroughly in `.notes/meeting_notes.md: Documentation of AI interactions.` with detailed explanations of prompts, responses, and decisions"
      ]
    },
    "safety_requirements": [
      "NEVER break type safety",
      "ALWAYS maintain proper error handling and logging mechanisms",
      "ALWAYS document new code, including docstrings and inline comments"
    ],
    "priorities": [
      {
        "source": ".notes/",
        "weight": 1.0
      }
    ],
    "modes": {
      "base": {
        "description": "For routine tasks that require standard deep learning development"
      },
      "enhanced": {
        "description": "For advanced tasks, such as multi-GPU distributed training or complex debugging scenarios"
      }
    },
    "project_directives": {
      "name": "my_project",
      "ai_first": true
    }
  }
}