config.json

{
  "architectures": [
    "NemotronForCausalLM"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "bos_token_id": 0,
  "eos_token_id": 1,
  "head_dim": 128,
  "hidden_act": "relu2",
  "hidden_size": 6144,
  "initializer_range": 0.008,
  "intermediate_size": 36864,
  "max_position_embeddings": 131072,
  "mlp_bias": false,
  "model_type": "nemotron",
  "norm_eps": 1e-05,
  "num_attention_heads": 48,
  "num_hidden_layers": 40,
  "num_key_value_heads": 8,
  "partial_rotary_factor": 0.5,
  "rope_theta": 2000000.0,
  "tie_word_embeddings": false,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.51.3",
  "use_cache": true,
  "vocab_size": 128000,
  "training_steps": 1170080,
  "training_tokens": 4907671224320
}