config.json

{
  "model_type": "sykollm",
  "architectures": [
    "SykoLLM"
  ],
  "class_name": "SykoLLM",
  "description": "Custom NanoGPT style decoder-only transformer model",
  "vocab_size": 8000,
  "n_embd": 64,
  "n_layer": 4,
  "n_head": 4,
  "block_size": 512,
  "max_position_embeddings": 512,
  "dropout": 0.0,
  "bias": true,
  "layer_norm_epsilon": 1e-05,
  "initializer_range": 0.02,
  "bos_token_id": 1,
  "eos_token_id": 2,
  "pad_token_id": 0,
  "unk_token_id": 3,
  "use_cache": true,
  "device": "cuda",
  "version": "1.0"
}