{
  "architecture": "nanoGPT (custom, trained from scratch)",
  "model_type": "classifier (binary spam detection)",
  "base_model": "nishantup/nanogpt-slm-124m",
  "model_config": {
    "block_size": 256,
    "vocab_size": 50257,
    "n_layer": 12,
    "n_head": 12,
    "n_embd": 768,
    "dropout": 0.0,
    "bias": true
  },
  "classifier_config": {
    "num_classes": 2,
    "labels": {
      "0": "not spam",
      "1": "spam"
    },
    "max_length": 120,
    "pad_token_id": 50256,
    "classification_head": "lm_head replaced with Linear(768, 2)"
  },
  "total_parameters_millions": 123.9,
  "tokenizer": "tiktoken gpt2 (50,257 BPE tokens)",
  "framework": "PyTorch",
  "training": {
    "dataset": "UCI SMS Spam Collection (balanced: 747 spam + 747 ham)",
    "split": "70% train / 10% test / 20% validation",
    "epochs": 5,
    "optimizer": "AdamW (lr=5e-5, weight_decay=0.1)",
    "max_length": 120,
    "frozen_layers": "All except last transformer block, final LayerNorm, and lm_head"
  }
}