zengxianyu / crfill

CR-Fill: Generative Image Inpainting with Auxiliary Contextual Reconstruction. ICCV 2021
Other
225 stars 43 forks source link

Issue with the model outputs at inference #7

Open hmrishavbandy opened 3 years ago

hmrishavbandy commented 3 years ago

Issue with the final stage of training :

python train.py \
    ${PREFIX} \
    --batchSize ${BSIZE0} \
    --nThreads ${NWK} \
    --update_part all \
    --niter 10 \
    ${EXTRA}

The issue I am facing:

Traceback (most recent call last):
  File "train.py", line 60, in <module>
    infer_out,inp = trainer.pix2pix_model.forward(data_i, mode='inference')
ValueError: too many values to unpack (expected 2)
hmrishavbandy commented 3 years ago

The hyperparameters for reference:

PREFIX="--dataset_mode_train trainimage \
--name debugarr0 \
--gpu_ids 0 \
--dataset_mode_val valimage \
--train_image_dir ... \
--train_image_list ...  \
--path_objectshape_base ... \
--path_objectshape_list ... \
--val_image_dir ... \
--val_image_list ... \
--val_mask_dir ... \
--no_vgg_loss \
--no_ganFeat_loss \
--load_size 256 \
--crop_size 256 \
--model arrange \
--netG twostagend \
--baseG baseconv \
--norm_type 1 \
--netD deepfill \
--load_base_g ... \
--load_base_d ... \
--lambda_ref 0.5 \
--lambda_l1 1 \
--preprocess_mode scale_shortside_and_crop"
zengxianyu commented 3 years ago

it's fixed. please pull to use the updated code

mac744mail commented 3 years ago

same error

hmrishavbandy commented 3 years ago

Works

hmrishavbandy commented 3 years ago

How many epochs did you run for each stage in the training? The epochs listed in the bash file seem too low. Did you use that low epochs?

zengxianyu commented 3 years ago

it depends on the size of training dataset. The epoch listed in the bash file is for Places (8 million images)