config.json

{
  "model_type": "multiscale_transformer",
  "architectures": [
    "MultiScaleForCausalLM"
  ],
  "vocab_size": 258,
  "d_model": 288,
  "n_heads": 4,
  "d_ff": 768,
  "n_layers_per_scale": 4,
  "n_cross_attn_layers": 1,
  "max_seq_len": 1024,
  "dropout": 0.0,
  "bias": false,
  "rope_theta": 10000.0,
  "downsample_factors": [
    1,
    2,
    4
  ],
  "num_parameters": 18090144,
  "training_results": {
    "model": "AXL-Refactor-20M",
    "params": 19098720,
    "steps": 202,
    "time": 300.3171327114105,
    "final_loss": 0.008088511414825916,
    "perplexity": 1.01,
    "max_seq_len": 1024,
    "context_window": "1024 bytes"
  }
}