hkust-knowcomp / exo-pcr Goto Github PK

Hello, thanks for sharing your great work.
Could this approach be used in an end-to-end manner or a different dataset?

Thanks in advance

Hi,
I am trying to reproduce your code but got an error:

File "./tensorflow/python/training/checkpoint_utils.py", line 229, in _init_from_checkpoint
    tensor_name_in_ckpt, str(variable_map[tensor_name_in_ckpt])

ValueError: Shape of variable coref_layer/slow_antecedent_scores/hidden_weights_0:0 ((7012, 3000)) doesn't match with shape of tensor coref_layer/slow_antecedent_scores/hidden_weights_0 ([7052, 3000]) from checkpoint reader.

Hi,

When I trained bert_base model with the train.py, it gave me such error after "Loaded 500 eval examples":

Traceback (most recent call last):
File "train.py", line 93, in
eval_summary, eval_f1 = model.evaluate(session, tf_global_step)
File "/Exo-PCR/independent.py", line 738, in evaluate
coref_predictions[example["doc_key"]] = self.evaluate_coref(top_span_starts, top_span_ends, predicted_antecedents, gold_clusters, coref_evaluator)
File "/Exo-PCR/independent.py", line 696, in evaluate_coref
evaluator.update(predicted_clusters, gold_clusters, mention_to_predicted, mention_to_gold)
File "/Exo-PCR/metrics.py", line 23, in update
e.update(predicted, gold, mention_to_predicted, mention_to_gold)
File "/Exo-PCR/metrics.py", line 48, in update
pn, pd, rn, rd = self.metric(predicted, gold)
File "/Exo-PCR/metrics.py", line 128, in ceafe
similarity = sum(scores[matching[:, 0], matching[:, 1]])
TypeError: tuple indices must be integers or slices, not tuple

The logs look fine in TensorBoard. Have you encountered a similar error on your end? Could you please help me take a look what may cause the error?

Thanks!