{ "architectures": [ "AblationForCausalLM" ], "attention_type": "standard", "auto_map": { "AutoConfig": "configuration_ablation.AblationConfig", "AutoModelForCausalLM": "modeling_ablation.AblationForCausalLM" }, "dsa_index_dim": 16, "dsa_index_heads": 2, "dsa_topk": 512, "dtype": "float32", "eos_token_id": 0, "hidden_size": 128, "initializer_range": 0.02, "intermediate_size": 336, "max_position_embeddings": 3072, "model_type": "ablation_lm", "num_attention_heads": 4, "num_hidden_layers": 5, "pad_token_id": 0, "rms_norm_eps": 1e-05, "rope_base": 10000.0, "transformers_version": "5.14.1", "vocab_size": 4096 }