config_standard.yaml

# StackWise Configuration - Standard Tokenizer Version
# Hierarchical configuration with validation and defaults
# Compatible with standard tokenizers (GPT-2, BERT, etc.)

model:
  # Model dimensions - Standard sizes for compatibility
  vocab_size: 50257  # GPT-2 vocabulary size
  d_model: 768
  n_heads: 12
  n_kv_heads: 12
  d_ff: 3072
  
  # Architecture configuration
  architecture:
    n_stacks: 2        # Number of stacks (groups of blocks)
    blocks_per_stack: 6  # Number of blocks per stack (12 total blocks)
    
  # Attention configuration
  attention_preset: "gpt_style"    # bert_style | gpt_style | efficient_gqa | mla_attention | kernel_attention | mlgka | custom
  
  # Custom attention configuration (only used when attention_preset: "custom")
  attention_custom:
    attention_type: "mha"              # mha | mla (GQA is determined by n_kv_heads)
    attention_mode: "causal"           # bidirectional | causal
    
    # MLA specific parameters
    mla_rq: 64
    mla_rkv: 32
    
    # Kernel attention parameters
    kernel_type: "linear"             # linear | gaussian | laplacian | uniform
    kernel_dim: 32
  
  # Normalization and MLP
  dropout: 0.1
  tie_embeddings: true
  freeze_up_proj: false
  
  # Positional encoding
  use_rope: true
  rope_theta: 10000.0
  
  # Mask-diffusion parameters
  mask_fraction_min: 0.15
  mask_fraction_max: 0.90
  special_mask_id: 4
  
  # Tokenizer and embedding configuration
  tokenizer_embedding:
    family: "gpt2"
    embedding_option: "embed_tokens"
    freeze_embeddings: false
    adapter_hidden_dim: null

training:
  # Training parameters
  batch_size: 8
  seq_len: 512
  max_steps: 100
  
  # Optimizer configuration
  optimizer:
    optimizer_type: "AdamW"  # AdamW | SGD | custom
    lr: 5.0e-5
    weight_decay: 0.01
    betas: [0.9, 0.95]
    eps: 1.0e-8
    # SGD specific (ignored for AdamW)
    momentum: 0.9
    dampening: 0.0
    nesterov: false
    # Custom parameters for advanced users
    custom_params: {}
    # Grouped parameters (optional)
    use_groups: false
    groups: []
  
  # Device and memory
  device: "cuda"
  gradient_checkpointing: false
  
  # Caching configuration
  cache_dir: "./cache"
  cache_mode: "stack"              # stack | rack
  
  # Saving configuration
  save_stacks: true                # Always save individual stacks (default enabled)
  save_rack: false                 # Optionally save entire rack (default disabled)
  
  # Mask-diffusion training
  min_mask_fraction: 0.15
  max_mask_fraction: 0.90
  mask_schedule_type: "linear"      # linear | exponential | cosine
  mask_token_id: null  # Will be set from tokenizer
  epochs_per_stack: 1
  
  # Training strategy: HOW to train
  strategy: "progressive"           # "progressive" | "end_to_end"
  # progressive: Build and train stacks one by one
  # end_to_end: Train the entire model at once
  
  # End-to-end training scope: WHAT to train (only used when strategy="end_to_end")
  end_to_end_scope: "stackwise"    # "stackwise" | "rackwise"
  # stackwise: Train each stack independently by caching previous stack activations
  # rackwise: Train the entire rack together
  
  # Progressive training configuration
  progressive:
    enabled: true
    max_stacks: 12
    target_stacks: 6                # Number of stacks to build progressively
    building_mode: "append"         # "append" or "prepend"
    trunk_strategy: "frozen"        # "frozen" | "qlora"
    new_stack_precision: "full"     # "full" | "nf_fp8" | "fp16"
    cache_activations: true
    time_interpretation: "depth"    # "depth" or "input"
    training_objective: "clm"       # "mlm" | "clm" | "custom"
  
  # Fine-tuning
  fine_tune_mode: "clm"             # clm | mlm | diffusion
  
  # Run identification and organization
  run_id: "standard_run"            # Unique identifier for this training run
  
  # QLoRA and quantization settings
  qlora:
    enabled: true
    rank: 16                        # Rank of low-rank adaptation matrices
    alpha: 32                       # Scaling factor (typically 2x rank)
    dropout: 0.1                    # Dropout rate for QLoRA adapters
    lr: 1.0e-5                     # Learning rate for QLoRA parameters
    
    # Progressive QLoRA settings
    progressive_enabled: false
    progressive_rank: 8             # Base rank for progressive strategies
    progressive_alpha: 16           # Base alpha for progressive strategies
    
    # QLoRA Strategy and Patterns
    strategy: "simplified"          # simplified | progressive | variable
    # simplified: Same QLoRA config for all stacks
    # progressive: Gradually change QLoRA parameters across stacks
    # variable: Custom QLoRA config per stack (defined in configs dict)
    
    # Mixed precision training (auto-enabled with QLoRA)
    mixed_precision: true           # Frozen trunk in fp16 + adapters in fp32
    
    rank_pattern: "constant"        # constant | increasing | decreasing
    # constant: Same rank for all stacks
    # increasing: Rank increases with stack depth (early: low rank, later: high rank)
    # decreasing: Rank decreases with stack depth (early: high rank, later: low rank)
    
    alpha_pattern: "constant"       # constant | increasing | decreasing
    # constant: Same alpha for all stacks
    # increasing: Alpha increases with stack depth
    # decreasing: Alpha decreases with stack depth
  current_block_lr: 5.0e-5
  quantization_enabled: true
  quantization_type: "fp16"         # fp4 | fp8 | fp16 | fp32
  
  # Time-step-based masking
  time_step_masking: true           # Enable time-step-based masking
  num_time_steps: 6                 # Number of discrete time steps
  time_step_mask_fractions: [0.15, 0.25, 0.35, 0.45, 0.55, 0.65]
  
  # Progressive training configuration
  progressive:
    enabled: true                   # Enable progressive training
    trunk_strategy: "frozen"        # "frozen" or "qlora"
    new_stack_precision: "full"     # "full", "half", "bfloat16", "qlora"
    cache_activations: true         # Cache activations for trunk training
    
    # Note: QLoRA configuration is now in the main qlora section above
    
    max_stacks: 12                  # Maximum number of stacks
    
    # Training objective
    training_objective: "clm"       # "mlm", "clm", "custom"
    
    # Time interpretation
    time_interpretation: "depth"    # "input" or "depth"
  
  # Logging and checkpointing
  log_interval: 10
  save_interval: 100
  checkpoint_dir: "./checkpoints"

data:
  # Dataset parameters
  dataset_path: null
  use_dummy_data: true
  num_samples: 1000
  
  # Data preprocessing
  tokenizer_path: null
  max_length: 512
  padding: "right"
  
  # Data loading
  num_workers: 0
  pin_memory: true
  shuffle: true