CoolFace
Modelpublic

adsingh64/gemma_sleeper_agent

sourceHugging Faceupdated 1y agoView on Hugging Face
0likes8downloads
config.json33 linesDownload Raw Back to root
1{2  "architectures": [3    "Gemma2ForCausalLM"4  ],5  "attention_bias": false,6  "attention_dropout": 0.0,7  "attn_logit_softcapping": 50.0,8  "bos_token_id": 256000,9  "cache_implementation": "hybrid",10  "eos_token_id": 256001,11  "final_logit_softcapping": 30.0,12  "head_dim": 256,13  "hidden_act": "gelu_pytorch_tanh",14  "hidden_activation": "gelu_pytorch_tanh",15  "hidden_size": 2304,16  "initializer_range": 0.02,17  "intermediate_size": 9216,18  "max_position_embeddings": 8192,19  "model_type": "gemma2",20  "num_attention_heads": 8,21  "num_hidden_layers": 26,22  "num_key_value_heads": 4,23  "pad_token_id": 256001,24  "query_pre_attn_scalar": 256,25  "rms_norm_eps": 1e-06,26  "rope_theta": 10000.0,27  "sliding_window": 4096,28  "torch_dtype": "float32",29  "transformers_version": "4.52.4",30  "use_cache": true,31  "vocab_size": 25600232}33