{
  "model_type": "seqcond",
  "architectures": [
    "SeqCondForCausalLM"
  ],
  "auto_map": {
    "AutoConfig": "configuration_seqcond.SeqCondConfig",
    "AutoModelForCausalLM": "modeling_seqcond.SeqCondForCausalLM",
    "AutoTokenizer": "tokenization_seqcond.SeqCondTokenizer"
  },
  "transformers_version": "5.3.0",
  "d_model": 1024,
  "d_ff": 2730,
  "num_layers": 24,
  "vocab_size": 100300,
  "maxlen": 4096,
  "num_heads": 16,
  "num_kv_heads": 4,
  "qk_norm": true,
  "qk_norm_eps": 1e-06,
  "seqcond_heads": 16,
  "num_query_heads": 16,
  "num_thetas": 2,
  "conv_kernel_size": 4,
  "expand_factor": 2.0,
  "out_expand_factor": 3,
  "seqcond_ratio": 2,
  "skip_low_rank": false,
  "num_anchor_heads": 0,
  "eos_token_id": 100279,
  "pad_token_id": 100279,
  "bos_token_id": null
}