INNER CODE UNIT · Python

end

ShannonAI/glyce · glyce/models/latticeLSTM/main.py:297

            end = min((batch_id+1)*data.HP_batch_size, train_num)
            instance = data.train_Ids[start:end]
            if not instance:
                continue
            gaz_list,  batch_word, batch_biword, batch_wordlen, batch_wordrecover, batch_char, batch_charlen, batch_charrecover, batch_label, mask  = batchify_with_label(instance, data.HP_gpu)
            instance_count += 1
            loss, tag_seq = model.neg_log_likelihood_loss(gaz_list, batch_word, batch_biword, batch_wordlen, batch_char, batch_charlen, batch_charrecover, batch_label, mask)
            right, whole = predict_check(tag_seq, batch_label, mask)
            right_token += right
            whole_token += whole
            sample_loss += loss.data[0]
            total_loss += loss.data[0]
            batch_loss += loss

            if end % 500 == 0:
                temp_time = time.time()
                temp_cost = temp_time - temp_start
                temp_start = temp_time

View source record →

📰 Research Paper
Loading…
⏳ Fetching content…