| { | |
| "_name_or_path": "sundial_base_128m", | |
| "architectures": [ | |
| "SundialForPrediction" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "configuration_sundial.SundialConfig", | |
| "AutoModelForCausalLM": "modeling_sundial.SundialForPrediction" | |
| }, | |
| "diffusion_batch_mul": 4, | |
| "dropout_rate": 0.1, | |
| "flow_loss_depth": 3, | |
| "hidden_act": "silu", | |
| "hidden_size": 768, | |
| "initializer_range": 0.02, | |
| "input_token_len": 16, | |
| "intermediate_size": 3072, | |
| "max_position_embeddings": 10000, | |
| "model_type": "sundial", | |
| "num_attention_heads": 12, | |
| "num_hidden_layers": 12, | |
| "num_sampling_steps": 50, | |
| "output_token_lens": [ | |
| 720 | |
| ], | |
| "rope_theta": 10000, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.40.1", | |
| "use_cache": true | |
| } | |