{ "attention_bias": false, "attention_dropout": 0.0, "eos_token_id": 151645, "hidden_act": "silu", "hidden_size": 2048, "initializer_range": 0.02, "intermediate_size": 4864, "max_position_embeddings": 4096, "model_type": "rish_ai", "norm_topk_prob": false, "num_attention_heads": 16, "num_experts": 7, "num_experts_per_tok": 5, "num_hidden_layers": 24, "num_key_value_heads": 16, "output_router_logits": false, "pad_token_id": 151643, "rms_norm_eps": 1e-06, "rope_parameters": { "factor": 1.0, "rope_theta": 500000.0, "rope_type": "dynamic" }, "rope_theta": 500000.0, "router_aux_loss_coef": 0.01, "tie_word_embeddings": false, "transformers_version": "5.0.0rc1", "use_cache": true, "vocab_size": 151680 }