{ "architectures": ["ParamBharatGenForCausalLM"], "model_type": "parambharatgen", "auto_map":{ "AutoConfig": "config_parambharatgen.ParamBharatGenConfig", "AutoModelForCausalLM": "modeling_parambharatgen.ParamBharatGenForCausalLM" }, "rope_theta": 10000.0, "use_cache": true, "hidden_size": 4096, "max_position_embeddings": 4096, "attention_dropout": 0.0, "num_hidden_layers": 16, "intermediate_size": 14336, "num_attention_heads": 32, "initializer_range": 0.01, "rms_norm_eps": 1e-05, "pretraining_tp": 1, "num_key_value_heads": 8, "hidden_act": "silu", "tie_word_embeddings": false, "mlp_bias": false, "attention_bias": false, "vocab_size": 256000, "bos_token_id": 2, "eos_token_id": 3, "torch_dtype": "bfloat16", "transformers_version": "4.42.3" }