Abc7347/clone

1{2  "architectures": [3    "Gemma3ForCausalLM"4  ],5  "attention_bias": false,6  "attention_dropout": 0.0,7  "attn_logit_softcapping": null,8  "bos_token_id": 2,9  "cache_implementation": "hybrid",10  "eos_token_id": 106,11  "final_logit_softcapping": null,12  "head_dim": 128,13  "hidden_activation": "gelu_pytorch_tanh",14  "hidden_size": 5376,15  "initializer_range": 0.02,16  "intermediate_size": 21504,17  "max_position_embeddings": 131072,18  "model_type": "gemma3_text",19  "num_attention_heads": 32,20  "num_hidden_layers": 62,21  "num_key_value_heads": 16,22  "pad_token_id": 0,23  "query_pre_attn_scalar": 168,24  "rms_norm_eps": 1e-06,25  "rope_local_base_freq": 10000,26  "rope_scaling": {27    "factor": 8.0,28    "rope_type": "linear"29  },30  "rope_theta": 1000000,31  "sliding_window": 1024,32  "sliding_window_pattern": 6,33  "torch_dtype": "bfloat16",34  "transformers_version": "4.51.3",35  "unsloth_fixed": true,36  "use_cache": true,37  "vocab_size": 26214438}39