backbone_layers:
- 2
- 3
- 7
batchsize: 12
betas:
- 0.9
- 0.999
bos_token: 1
channels: 1
config: colab.yaml
data: dataset/data/simple_train.pkl
debug: false
decoder_args:
  attn_on_attn: true
  cross_attend: true
  ff_glu: true
  rel_pos_bias: false
  use_scalenorm: false
device: cuda:0
dim: 256
encoder_depth: 4
encoder_structure: hybrid
eos_token: 2
epoch: 429
epochs: 500
gamma: 0.9995
gpu_devices:
- 0
heads: 8
id: v9h46w6a
load_chkpt: /home/coder/project/LaTeX-OCR/weights.pth
lr: 0.001
lr_step: 30
max_dimensions:
- 336
- 336
max_height: 336
max_seq_len: 2048
max_width: 336
min_dimensions:
- 32
- 32
min_height: 32
min_width: 32
model_path: simple_checkpoints
name: mixed
no_cuda: false
num_layers: 4
num_tokens: 8000
optimizer: Adam
output_path: simple_outputs
pad: false
pad_token: 0
patch_size: 16
resume: false
sample_freq: 201
save_freq: 1
scheduler: StepLR
seed: 42
temperature: 0.2
test_samples: 5
testbatchsize: 8
tokenizer: dataset/tokenizer.json
valbatches: 1
valdata: dataset/data/simple_val.pkl
wandb: true