{
  "_name_or_path": "configs/gsa_16M.json",
  "architectures": [
    "GSAForCausalLM"
  ],
  "attn": null,
  "attn_mode": "chunk",
  "bos_token_id": 1,
  "clamp_max": null,
  "clamp_min": null,
  "conv_size": 4,
  "elementwise_affine": true,
  "eos_token_id": 2,
  "expand_k": 1,
  "expand_v": 1,
  "feature_map": "swish",
  "fuse_cross_entropy": true,
  "fuse_norm": true,
  "gate_logit_normalizer": 8,
  "hidden_act": "swish",
  "hidden_ratio": 4,
  "hidden_size": 256,
  "initializer_range": 0.02,
  "intermediate_size": null,
  "max_position_embeddings": 2048,
  "model_type": "gsa",
  "norm_eps": 1e-06,
  "norm_first": true,
  "num_heads": 4,
  "num_hidden_layers": 10,
  "num_kv_heads": null,
  "num_slots": 16,
  "tie_word_embeddings": true,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.47.0",
  "use_cache": true,
  "use_gk": true,
  "use_gv": false,
  "use_norm": true,
  "use_output_gate": false,
  "use_short_conv": false,
  "vocab_size": 32000
}