| { | |
| "architectures": [ | |
| "OBILanguageModel" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "modelConfig.OBIConfig", | |
| "AutoModel": "modelLM.OBILanguageModel", | |
| "AutoModelForCausalLM": "modelLM.OBILanguageModel", | |
| "AutoModelForQuestionAnswering": "modelLM.OBILanguageModel" | |
| }, | |
| "batch_size": 24, | |
| "block_size": 150, | |
| "device": "cuda", | |
| "eval_interval": 500, | |
| "hidden_dropout_prob": 0.1, | |
| "hidden_size": 256, | |
| "learning_rate": 0.001, | |
| "max_iters": 20000, | |
| "num_attention_heads": 32, | |
| "num_hidden_layers": 24, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.34.0", | |
| "vocab_size": 5000 | |
| } | |