total_params: 66889174
image_size: 128
num_tokens: 8192
num_train_steps: -1
num_epochs: 20
dim: 64
channels: 4
batch_size: 1
lr: 0.0001
gradient_accumulation_steps: 100
save_results_every: 500
save_model_every: 5000
vq_codebook_size: 8192
vq_codebook_dim: 8192
lr_scheduler: cosine_with_restarts
lr_warmup_steps: 100
seq_len: 8192
depth: 4
dim_head: 64
heads: 8
ff_mult: 4
t5_name: t5-small
mixed_precision: 'no'
cond_image_size: null
validation_prompt: a girl|a cat|a dog
timesteps: 18
optimizer: Adam
only_save_last_checkpoint: false
validation_image_scale: 1.0
no_center_crop: false
no_flip: false
dataset_save_path: E:\cached_datasets\INE2
clear_previous_experiments: false
max_grad_norm: null
seed: 42
valid_frac: 0.05
use_ema: false
ema_beta: 0.995
ema_update_after_step: 1
ema_update_every: 1
apply_grad_penalty_every: 4
image_column: image
caption_column: caption
log_with: wandb
use_8bit_adam: false
results_dir: results\Muse-v0.10
logging_dir: null
vae_path: results\Muse-v0.10\vae.1456000.pt
dataset_name: null
hf_split_name: null
streaming: false
train_data_dir: E:\dataset
checkpoint_limit: null
cond_drop_prob: 0.5
scheduler_power: 1.0
num_cycles: 200
resume_path: results\Muse-v0.10\maskgit.143000.pt
taming_model_path: null
taming_config_path: null
weight_decay: 0.0001
cache_path: null
no_cache: true
link: false
latest_checkpoint: true
do_not_save_config: false
use_l2_recon_loss: false
debug: false
config_path: null
attention_type: ein
random_crop: true
project_name: muse_maskgit
run_name: null
wandb_user: sygil
layers: 4
discr_layers: 4