state_dict_report.json

{
  "base_model": "bartpho",
  "total_params": 531,
  "expected_heads_present": {
    "cls_decoder.weight": false,
    "cls_decoder.bias": false,
    "cls_dense.weight": false,
    "cls_dense.bias": false,
    "cls_layer_norm.weight": false,
    "cls_layer_norm.bias": false,
    "mask_n_predictor.mask_predictor_dense.weight": true,
    "mask_n_predictor.mask_predictor_dense.bias": true,
    "mask_n_predictor.mask_predictor_proj.weight": true,
    "mask_n_predictor.mask_predictor_proj.bias": true,
    "nsw_detector.dense.weight": true,
    "nsw_detector.dense.bias": true,
    "nsw_detector.predictor.weight": true,
    "nsw_detector.predictor.bias": true
  },
  "alt_common_heads_present": {
    "lm_head.weight": false,
    "lm_head.bias": false,
    "cls.decoder.weight": true,
    "cls.decoder.bias": true,
    "cls.dense.weight": true,
    "cls.dense.bias": true,
    "cls.layer_norm.weight": true,
    "cls.layer_norm.bias": true
  },
  "aux_heads_present": {
    "nsw_detector.": true,
    "mask_n_predictor.": true
  },
  "example_keys": [
    "bart.shared.weight",
    "bart.encoder.embed_tokens.weight",
    "bart.encoder.embed_positions.weight",
    "bart.encoder.layers.0.self_attn.k_proj.weight",
    "bart.encoder.layers.0.self_attn.k_proj.bias",
    "bart.encoder.layers.0.self_attn.v_proj.weight",
    "bart.encoder.layers.0.self_attn.v_proj.bias",
    "bart.encoder.layers.0.self_attn.q_proj.weight",
    "bart.encoder.layers.0.self_attn.q_proj.bias",
    "bart.encoder.layers.0.self_attn.out_proj.weight",
    "bart.encoder.layers.0.self_attn.out_proj.bias",
    "bart.encoder.layers.0.self_attn_layer_norm.weight",
    "bart.encoder.layers.0.self_attn_layer_norm.bias",
    "bart.encoder.layers.0.fc1.weight",
    "bart.encoder.layers.0.fc1.bias",
    "bart.encoder.layers.0.fc2.weight",
    "bart.encoder.layers.0.fc2.bias",
    "bart.encoder.layers.0.final_layer_norm.weight",
    "bart.encoder.layers.0.final_layer_norm.bias",
    "bart.encoder.layers.1.self_attn.k_proj.weight"
  ]
}