config.json

{
  "architectures": [
    "GemmaForCausalLM"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "bos_token_id": 2,
  "dtype": "bfloat16",
  "eos_token_id": 1,
  "ft_cfg": {
    "bf16": true,
    "continue_final_message": true,
    "dataset_name": "eekay/gemma-2b-it-eagle-numbers",
    "gradient_accumulation_steps": 1,
    "learning_rate": 0.0002,
    "logging_steps": 100,
    "lora_rank": 8,
    "lr_scheduler_type": "constant",
    "max_grad_norm": 1.0,
    "model_id": "google/gemma-2b-it",
    "model_save_name": "gemma-2b-it-eagle-numbers-ft",
    "n_examples": 30000,
    "num_train_epochs": 1,
    "per_device_train_batch_size": 24
  },
  "head_dim": 256,
  "hidden_act": "gelu",
  "hidden_activation": null,
  "hidden_size": 2048,
  "initializer_range": 0.02,
  "intermediate_size": 16384,
  "max_position_embeddings": 8192,
  "model_type": "gemma",
  "num_attention_heads": 8,
  "num_hidden_layers": 18,
  "num_key_value_heads": 1,
  "pad_token_id": 0,
  "rms_norm_eps": 1e-06,
  "rope_scaling": null,
  "rope_theta": 10000.0,
  "transformers_version": "4.57.1",
  "use_cache": true,
  "vocab_size": 256000
}