| { | |
| "architectures": [ | |
| "YForCausalLM2" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "ymodel2.YConfig2", | |
| "AutoModelForCausalLM": "ymodel2.YForCausalLM2" | |
| }, | |
| "bos_token_id": 1, | |
| "dropout": 0.1, | |
| "eos_token_id": 2, | |
| "head_dim": 128, | |
| "hidden_act": "gelu_pytorch_tanh", | |
| "hidden_size": 768, | |
| "intermediate_size": 2048, | |
| "max_position_embeddings": 4096, | |
| "model_type": "ynet2", | |
| "num_heads": 16, | |
| "num_layers": 16, | |
| "rms_norm_eps": 1e-08, | |
| "rope_theta": 50000.0, | |
| "self_distill": true, | |
| "torch_dtype": "bfloat16", | |
| "transformers_version": "4.51.3", | |
| "vocab_size": 6400 | |
| } | |