| { | |
| "architectures": [ | |
| "SeedForCausalLM" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "configuration_seed.SeedConfig", | |
| "AutoModelForCausalLM": "modeling_seed.SeedForCausalLM" | |
| }, | |
| "model_type": "seed", | |
| "vocab_size": 64000, | |
| "n_embd": 1024, | |
| "n_layer": 28, | |
| "n_head": 16, | |
| "n_kv_head": 8, | |
| "head_dim": 128, | |
| "mlp_hidden_dim": 3072, | |
| "block_size": 4096, | |
| "bias": false, | |
| "dropout": 0.0, | |
| "rms_norm_eps": 1e-6, | |
| "rope_theta": 1000000.0, | |
| "rope_scaling_type": "none", | |
| "rope_scaling_factor": 1.0, | |
| "tie_word_embeddings": true, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.57.3" | |
| } | |