| { | |
| "model_type": "custom_diffusion", | |
| "architecture": "CogVideoX", | |
| "transformers_version": "4.41.0", | |
| "diffusion_config": { | |
| "num_timesteps": 1000, | |
| "beta_start": 0.0001, | |
| "beta_end": 0.02, | |
| "scheduler_type": "default", | |
| "prediction_type": "epsilon", | |
| "rescale_timesteps": true | |
| }, | |
| "model_config": { | |
| "num_hidden_layers": 12, | |
| "num_attention_heads": 16, | |
| "hidden_size": 768, | |
| "intermediate_size": 3072, | |
| "max_position_embeddings": 512, | |
| "vocab_size": 49408, | |
| "hidden_act": "gelu", | |
| "initializer_range": 0.02 | |
| }, | |
| "training_config": { | |
| "learning_rate": 5e-5, | |
| "batch_size": 8, | |
| "num_train_steps": 100000 | |
| }, | |
| "pipeline_config": { | |
| "pipeline_type": "text-to-video", | |
| "input_format": "text", | |
| "output_format": "video" | |
| } | |
| } | |