Some of the important settings I used:
"num_of_dataset_images": 57,
"num_vectors_per_token": 1,
"embedding_name": "juliatica",
"learn_rate": "0.05:10, 0.02:20, 0.01:60, 0.005:1000, 0.002:2000, 0.001:35000, 0.0005",
"gradient_step": 1,
"steps": 60000,
"clip_grad_value": "0.1",
"latent_sampling_method": "deterministic",
Some of the important settings I used:
"num_of_dataset_images": 57,
"num_vectors_per_token": 1,
"embedding_name": "juliatica",
"learn_rate": "0.05:10, 0.02:20, 0.01:60, 0.005:1000, 0.002:2000, 0.001:35000, 0.0005",
"gradient_step": 1,
"steps": 60000,
"clip_grad_value": "0.1",
"latent_sampling_method": "deterministic",