{ "architectures": [ "Mahler60/Prueba" ], "attention_probs_dropout_prob": 0.1, "bos_token_id": 0, "eos_token_id": 1, "hidden_act": "gelu_new", "initializer_range": 0.02, "intermediate_size": 11008, "layer_norm_eps": 1e-5, "max_position_embeddings": 2048, "model_type": "gpt_neox", "n_heads": 32, "n_layers": 20, "n_embd": 2560, "vocab_size": 50257, "pad_token_id": 0, "embedding_size": 2560 }