hand-read2016-page / config.json
MHamdan's picture
Release weights (HAND-Decoding v1.1.0)
eacef65 verified
Raw History Blame Contribute Delete
1.35 kB
{
"model_type": "hand-dan-page",
"input_channels": 3,
"dropout": 0.5,
"enc_dim": 256,
"nb_layers": 5,
"pe_h_max": 500,
"pe_w_max": 1000,
"l_max": 15000,
"dec_num_layers": 8,
"dec_num_heads": 4,
"dec_res_dropout": 0.1,
"dec_pred_dropout": 0.1,
"dec_att_dropout": 0.1,
"dec_dim_feedforward": 256,
"attention_win": 100,
"use_tokens_from_all_lines": true,
"use_first_pass_tokens": true,
"two_step_pos_enc_mode": "cat",
"use_line_indices": false,
"vocab_size": 99,
"additional_tokens": 1,
"max_char_prediction": 3000,
"max_line_pred": 100,
"max_pred_per_line": 150,
"working_dpi": 150,
"image_mean": [
202.70203512453844,
190.85819291654516,
136.08993740340918
],
"image_std": [
71.95683449787683,
72.00679752035164,
57.33656372156373
],
"expected_parameters": 7033700,
"parameters_encoder": 1706240,
"parameters_decoder": 5327460,
"source_checkpoint": {
"path": "outputs/e14_budget_1p26M_s0/checkpoints/best_3580.pt",
"sha256": "d94af1419554a057ae671733af2fe4cf3f63d1f8b3652f9de17277b432632f5f",
"bytes": 84821823,
"epoch": 3580,
"step": 1253350,
"recorded_best_valid_cer": 0.0409,
"use_line_indices_source": "checkpoint carries no 'use_line_indices' key; inferred from additional_tokens == 1, i.e. (additional_tokens == 3) -> False"
}
}