config.json

{
  "architectures": [
    "OpensciForCausalLM"
  ],
  "attention_bias": true,
  "attention_dropout": 0.0,
  "auto_map": {
    "AutoConfig": "configuration_opensci.OpensciConfig",
    "AutoModel": "modeling_opensci.OpensciPreTrainedModel",
    "AutoModelForCausalLM": "modeling_opensci.OpensciForCausalLM"
  },
  "bos_token_id": 0,
  "dtype": "bfloat16",
  "eos_token_id": 0,
  "head_dim": 64,
  "hidden_act": "silu",
  "hidden_size": 1024,
  "initializer_range": 0.02,
  "intermediate_size": 3840,
  "layer_norm_eps": 1e-05,
  "max_position_embeddings": 4096,
  "mlp_bias": true,
  "model_type": "opensci",
  "num_attention_heads": 16,
  "num_hidden_layers": 22,
  "num_key_value_heads": 16,
  "pad_token_id": null,
  "pretraining_tp": 1,
  "qk_layernorm": true,
  "rms_norm_eps": 1e-05,
  "rope_parameters": null,
  "rope_theta": 100000,
  "tie_word_embeddings": true,
  "transformers_version": "5.3.0",
  "use_cache": true,
  "vocab_size": 50304
}