hf-tiny-model-private/tiny-random-GPTNeoForCausalLM

1{2  "activation_function": "gelu_new",3  "architectures": [4    "GPTNeoForCausalLM"5  ],6  "attention_dropout": 0.0,7  "attention_layers": [8    "global",9    "local",10    "global",11    "local"12  ],13  "attention_types": [14    [15      [16        "global",17        "local"18      ],19      220    ]21  ],22  "bos_token_id": 0,23  "embed_dropout": 0.0,24  "eos_token_id": 0,25  "hidden_size": 32,26  "initializer_range": 0.02,27  "intermediate_size": null,28  "is_decoder": true,29  "layer_norm_epsilon": 1e-05,30  "max_position_embeddings": 512,31  "model_type": "gpt_neo",32  "num_heads": 4,33  "num_layers": 4,34  "pad_token_id": 1023,35  "resid_dropout": 0.0,36  "torch_dtype": "float32",37  "transformers_version": "4.28.0.dev0",38  "use_cache": true,39  "vocab_size": 1024,40  "window_size": 741}42