| { |
| "architectures": [ |
| "Ernie4_5_ForCausalLM" |
| ], |
| "attention_probs_dropout_prob": 0.0, |
| "auto_map": { |
| "AutoConfig": "configuration_ernie4_5.Ernie4_5_Config", |
| "AutoModel": "modeling_ernie4_5.Ernie4_5_Model", |
| "AutoModelForCausalLM": "modeling_ernie4_5.Ernie4_5_ForCausalLM" |
| }, |
| "bos_token_id": 1, |
| "compression_ratio": 1.0, |
| "eos_token_id": 2, |
| "head_dim": 8, |
| "hidden_act": "silu", |
| "hidden_dropout_prob": 0.0, |
| "hidden_size": 32, |
| "ignored_index": -100, |
| "intermediate_size": 128, |
| "max_position_embeddings": 131072, |
| "max_sequence_length": null, |
| "model_type": "ernie4_5", |
| "num_attention_heads": 4, |
| "num_hidden_layers": 2, |
| "num_key_value_heads": 2, |
| "pad_token_id": 0, |
| "rms_norm_eps": 1e-05, |
| "rope_theta": 500000, |
| "tie_word_embeddings": true, |
| "torch_dtype": "float32", |
| "transformers_version": "4.52.0.dev0", |
| "use_bias": false, |
| "use_cache": false, |
| "use_flash_attention": false, |
| "vocab_size": 103424, |
| "weight_share_add_bias": true |
| } |
|
|
