config.json

{
  "alpha": 1.0,
  "attention_dropout": 0.0,
  "auto_map": {
    "AutoConfig": "configuration.ShramConfig",
    "AutoModelForCausalLM": "huggingface.ShramForCausalLM"
  },
  "beta": 32.0,
  "embedding_width": 512,
  "head_dim": 16,
  "inference_sequence_length": 1024,
  "local_rope_theta": 10000.0,
  "mlp_width": 1366,
  "model_type": "shram",
  "mosrah_rope_theta": 10000.0,
  "num_decoder_layers": 12,
  "num_mosrah_heads": 16,
  "num_selected_heads": 16,
  "num_sliding_window_heads": 16,
  "rms_norm_eps": 1e-05,
  "rope_mode": "main_sequence",
  "tie_word_embeddings": false,
  "training_sequence_length": 1024,
  "transformers_version": "5.12.1",
  "use_cache": true,
  "use_residual_gate": true,
  "vocab_size": 50277,
  "window_size": 128
}