| { | |
| "architectures": [ | |
| "WiolaForCausalLM" | |
| ], | |
| "atm_enabled": true, | |
| "atm_threshold": 0.92, | |
| "bos_token_id": 1, | |
| "dsff_narrow_size": 1024, | |
| "dsff_wide_size": 4096, | |
| "dtype": "float32", | |
| "eos_token_id": 2, | |
| "gcla_gate_init": -3.0, | |
| "gcla_lookback": 2, | |
| "hidden_size": 1024, | |
| "initializer_range": 0.02, | |
| "max_position_embeddings": 2048, | |
| "model_type": "wiola", | |
| "num_attention_heads": 16, | |
| "num_hidden_layers": 16, | |
| "num_key_value_heads": 4, | |
| "pad_token_id": 0, | |
| "rms_norm_eps": 1e-06, | |
| "srpe_radial_amplitude": 0.1, | |
| "srpe_radial_frequency": 0.01, | |
| "srpe_spiral_divisor": 8, | |
| "srpe_theta": 10000.0, | |
| "transformers_version": "4.57.6", | |
| "use_cache": true, | |
| "vocab_size": 32000, | |
| "auto_map": { | |
| "AutoConfig": "configuration_wiola.WiolaConfig", | |
| "AutoModel": "modeling_wiola.WiolaModel", | |
| "AutoModelForCausalLM": "modeling_wiola.WiolaForCausalLM" | |
| } | |
| } |