{
  "model_type": "alpha-er",
  "architectures": [
    "AlphaErForCausalLM"
  ],
  "auto_map": {
    "AutoModelForCausalLM": "modeling_alpha.AlphaErForCausalLM"
  },
  "vocab_size": 12288,
  "block_size": 512,
  "n_layer": 2,
  "n_embd": 1024,
  "n_head": 8,
  "ffn_dim": 20480,
  "mlp_experts": 64,
  "lm_head_rank": 128,
  "attn_rank": 128,
  "attn_soft_cap": 30.0,
  "layer_norm_eps": 1e-05,
  "torch_dtype": "float32",
  "trained_step": 20000
}