Potpov's picture
Upload mnist_bw/vsq/config.yaml with huggingface_hub
237e9e9 verified
Raw History Blame
2.25 kB
model_params:
name: "VSQ" # always keep this "VectorGPT"
vector_decoder_model: "cnn" # "mlp" or "raster_conv"
quantized_dim: 512
codebook_size: 4096 # will be ignored for FSQ
image_loss: "pyramid" # "pyramid" or "mse"
single_code_representation: true
vq_method: "fsq" # "vqvae", "FSQ", "vqtorch"
fsq_levels: [7,5,5,5,5] # will determine codebook_size, see Table 1 of FSQ paper - [7,5,5,5,5] for 4096 (e.g. StrokeNUWA), [8,5,5,5] for 1024
num_segments: 15 # default 15 - or 64
pred_color: false
data_params:
dataset: "mnistPrecomputed"
data_path: "/raid/marco.cipriano/data/SVG/Grimoire/MNIST/mnist_pretiled/P128_T6_P20_TH0.1"
train_batch_size: 2048
val_batch_size: 64
patch_size: 128 # HAS TO BE 224 FOR CLIP RES50 TO WORK, otherwise take 128
num_workers: 16
num_tiles_per_row: 6
random_colors: false
use_palette: false # use a color palette of 9 colors rather than random colors
padding_frac: 0.1 # fraction of the image that is padded with white background
exp_params:
lr: 0.00002 # 0.00002
weight_decay: 1.e-4 # specify positive float to enable, start experimenting with 1.e-4/1.e-3
scheduler_gamma: 0.98 # 0.95 is a good starting value
train_log_interval: 0.025 # len(dataset) / train_batch_size / desired_logging_frequency
manual_seed: 1265
schedule_pyramid_method: "linear" # "linear" or "exponential" or null to disable scheduling, default disabled
trainer_params:
devices: -1 # always keep at -1 as this takes all available GPUs specified through CUDA_VISIBLE_DEVICES
max_epochs: 250 # dsnt matter too much, got early stopping implemented
# accumulate_grad_batches: 2
logging_params:
entity: "aiis-chair" # comment to use default wandb entity "aiis-chair"
project: "grimoire-2" # your wandb project name
save_dir: "/raid/marco.cipriano/results/svg/Grimoire/VSQ"
# name: "VSQ_MNIST_BW_P128_T14_P20_TH0.2_S64" # name of the run in wandb
name: "SQ_MNIST_BW_P128_T6_P20_TH0.1" # name of the run in wandb
version: 1
author: "Marco" # will be a tag in wandb
# id: null # id of wandb run to continue
# allow_val_change: False # allow changing values in this config w.r.t. the run that you're continuing (good for changing loss weightings mid-run)