config/with_decoder.yaml

# config.yaml

exp_name: exp1-2
wandb: 90788c79e1500570b08e5acf283e17df7e0c54b2
root: "${oc.env:DATA_ROOT}"
overfit: False
batch_size: 4
num_workers: 4
img_size: 224
rationale_type: 0 # 0 - only rationale, 1 - randomly add entity desc, 2 - add entity desc
val_rationale_type: 0
hide_true_bbox: 8 # clues and inferences selected randomly
widescreen_processing: 1 # 0 - no widescreen, 1 - widescreen
h_flip: False
ema_decay: 0.9999
aux_weight: 0.2
no_hard_negative_itm: False

clip_model: 'ViT-B/16' # 'RN101' 'RN50x4''RN50x16' 'RN50x64' 'ViT-L/14@336px' 'ViT-B/32'
has_extra_txt_decoder: False
has_extra_img_decoder: False
has_extra_mix_decoder: False
has_extra_gen_decoder: False

extra_decoder:
  is_decoder: True
  vocab_size: 1000
  d_ff: 512
  d_kv: 64
  d_model: 512
  dropout_rate: 0.1
  num_heads: 8
  num_layers: 2
  # eos_token_id: 1
  # pad_token_id: 0
  # decoder_start_token_id: 0
  # n_positions: 512
  relative_attention_max_distance: 128
  relative_attention_num_buckets: 32

warmup: 1000
init_from: ''
lr: .00001
n_epochs: 15
save_every: 0
early_stop: 5
val_stat: 'loss'
device: 'cuda'
use_multi: False
local_rank: 0

hydra:
  run:
    dir: ./results/${exp_name}
  output_subdir: ./ # directory for saving the yaml configs
  job:
    config:
      override_dirname:
        exclude_keys:
          - exp.name