| pretrained_model_path: models/StableDiffusion/ | |
| motion_module: models/Motion_Module/mm_sd_v15.ckpt | |
| output_dir: models/Motion_Module/test | |
| train_data: | |
| video_folder: data/output | |
| csv_path: data/output.csv | |
| sample_n_frames: 5 | |
| sample_stride: 100 | |
| sample_size: 256 | |
| validation_data: | |
| prompts: | |
| - camera panning right to left, a bird's eye view of a row of buildings in a city | |
| with trees in the foreground, masterpiece, best quality | |
| video_length: 5 | |
| width: 256 | |
| height: 256 | |
| num_inference_steps: 20 | |
| guidance_scale: 12.5 | |
| temporal_context: 24 | |
| use_inv_latent: true | |
| num_inv_steps: 50 | |
| learning_rate: 0.0003 | |
| train_batch_size: 1 | |
| max_train_steps: 1 | |
| checkpointing_steps: 100 | |
| validation_steps: 1 | |
| trainable_modules: | |
| - to_q | |
| seed: 33 | |
| mixed_precision: fp16 | |
| use_8bit_adam: false | |
| gradient_checkpointing: true | |
| enable_xformers_memory_efficient_attention: true | |