{
  "architectures": [
    "QTensorFormerForCausalLM"
  ],
  "auto_map": {
    "AutoConfig": "configuration_qtensorformer.QTensorFormerConfig",
    "AutoModelForCausalLM": "modeling_qtensorformer.QTensorFormerForCausalLM"
  },
  "attention_sink_size": 4,
  "enable_dynamic_routing": true,
  "hidden_size": 2048,
  "hysteresis_threshold": 0.15,
  "initializer_range": 0.02,
  "intermediate_size": 5632,
  "kv_cache_window_size": 1024,
  "kv_quant_bits": 4,
  "max_position_embeddings": 4096,
  "model_type": "qtensorformer",
  "num_attention_heads": 16,
  "num_hidden_layers": 16,
  "num_key_value_heads": 4,
  "pid_kd": 0.05,
  "pid_ki": 0.01,
  "pid_kp": 0.1,
  "quantize_kv_cold": true,
  "rms_norm_eps": 1e-06,
  "rope_theta": 10000.0,
  "target_energy_budget": 1.0,
  "tie_word_embeddings": false,
  "transformers_version": "4.41.2",
  "tt_ranks": [
    1,
    16,
    16,
    1
  ],
  "use_cache": true,
  "vocab_size": 32000
}