| { | |
| "architectures": [ | |
| "HybridTimeScaleLM" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "model.HybridTimeScaleConfig", | |
| "AutoModelForCausalLM": "model.HybridTimeScaleLM" | |
| }, | |
| "bos_token_id": 1, | |
| "chunk_size": 128, | |
| "dropout": 0.0, | |
| "dtype": "bfloat16", | |
| "eos_token_id": 2, | |
| "latent_dim": 1344, | |
| "layer_types": [ | |
| "linear", | |
| "linear", | |
| "linear", | |
| "softmax", | |
| "linear", | |
| "linear", | |
| "linear", | |
| "softmax", | |
| "linear", | |
| "linear", | |
| "linear", | |
| "softmax", | |
| "linear", | |
| "linear", | |
| "linear", | |
| "softmax", | |
| "linear", | |
| "linear", | |
| "linear", | |
| "softmax", | |
| "linear", | |
| "linear", | |
| "linear", | |
| "softmax", | |
| "linear", | |
| "linear", | |
| "linear", | |
| "softmax", | |
| "linear", | |
| "linear", | |
| "linear", | |
| "softmax" | |
| ], | |
| "model_type": "hybrid_timescale_lm", | |
| "num_layers": 32, | |
| "num_modes": 16, | |
| "pad_token_id": 0, | |
| "time_scale": 128.0, | |
| "transformers_version": "4.56.0", | |
| "vocab_size": 151665 | |
| } | |