| { | |
| "architectures": [ | |
| "VMiniBertForMaskedLM" | |
| ], | |
| "attention_probs_dropout_prob": 0.1, | |
| "bos_token_id": 2, | |
| "dtype": "float32", | |
| "eos_token_id": 3, | |
| "hidden_act": "gelu", | |
| "hidden_dropout_prob": 0.1, | |
| "hidden_size": 768, | |
| "intermediate_size": 2048, | |
| "layer_norm_eps": 1e-06, | |
| "max_position_embeddings": 32, | |
| "model_type": "vminibert", | |
| "num_attention_heads": 12, | |
| "num_hidden_layers": 12, | |
| "pad_token_id": 0, | |
| "tie_word_embeddings": true, | |
| "transformers_version": "5.14.1", | |
| "vocab_size": 30000, | |
| "auto_map": { | |
| "AutoConfig": "configuration_vminibert.VMiniBertConfig", | |
| "AutoModelForMaskedLM": "modeling_vminibert.VMiniBertForMaskedLM" | |
| }, | |
| "total_params": 108615984, | |
| "test_top1_accuracy": "42.82%", | |
| "test_top3_accuracy": "53.55%", | |
| "test_top5_accuracy": "58.06%", | |
| "test_loss": 4.0378, | |
| "test_ppl": 56.7 | |
| } |