{ "architectures": [ "MobilintLlamaForCausalLM" ], "attention_bias": false, "attention_dropout": 0.0, "auto_map": { "AutoConfig": "proxy_llama.MobilintLlamaConfig", "AutoModelForCausalLM": "proxy_llama.MobilintLlamaForCausalLM" }, "bos_token_id": 100275, "end_token_id": 100275, "eos_token_id": 100275, "head_dim": 128, "hidden_act": "silu", "hidden_size": 2048, "initializer_range": 0.02, "intermediate_size": 7168, "max_position_embeddings": 4096, "mlp_bias": false, "model_type": "mobilint-llama", "mxq_path": "HyperCLOVAX-SEED-Text-Instruct-1.5B-Batch16-W4V8.mxq", "num_attention_heads": 16, "num_hidden_layers": 24, "num_key_value_heads": 8, "pad_token_id": 100275, "pretraining_tp": 1, "resid_pdrop": 0.2, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 100000000, "target_cores": [ "0:0" ], "tie_word_embeddings": false, "torch_dtype": "bfloat16", "transformers_version": "4.48.2", "use_cache": true, "vocab_size": 110592, "npu_prefill_chunk_size": { "single": 128 }, "max_batch_size": 16 }