{
  "architectures": [
    "OLM3NanoForCausalLM"
  ],
  "model_type": "olm3_nano",
  "vocab_size": 50304,
  "hidden_size": 2048,
  "intermediate_size": 5504,
  "num_hidden_layers": 16,
  "num_attention_heads": 16,
  "max_position_embeddings": 2048,
  "rope_theta": 10000.0,
  "rms_norm_eps": 1e-05,
  "torch_dtype": "float32",
  "transformers_version": "custom",
  "auto_map": {
    "AutoConfig": "configuration_olm3nano.OLM3NanoConfig",
    "AutoModelForCausalLM": "modeling_olm3nano.OLM3NanoForCausalLM"
  },
  "olm3_nano": {
    "checkpoint_step": 14086,
    "parameter_count": 1015613440,
    "training_corpus": "FineWeb"
  },
  "tie_word_embeddings": true
}