ardeal

I have a trained model, but the performance is not as good as that on the paper. I will re-train a better model and then upload it.

Hi @tucan9389 Many many thanks to you for your reply! I have already found those links you mentioned. The further questions are: 1) If I apply `q_aware_model = quantize_model(model)` to...

> Most likely the models that you are using use operators that are not implemented. Note that this project is not currently actively maintained. @alrevuelta as you said, this project...

1） 2080Ti， 如果batchsize=32，好像显存不够用。我的也是2080Ti，我设置的bathsize=4，好像设置batchsize=8也可以。16/32，我就没有尝试过。 2）149行出错，原因是evaluate返回的AP可能是空的。AP返回空的原因，可能是训练不好，导致evaluate效果很差。

我evaluate 时间很短，不需要十几分钟。 你是不是修改了什么，或者，你验证的样本数量是不是太多了？或者，你验证期间，是不是imshow了？ 你应该调试一下看看，到底是哪里慢。

You should prepare training and prediction data list. Modify the file config/coco.data. refer to the section how to run the training code.

@sngyai `Sequoia\strategy\backtrace_ma250.py'的第72行，有bug，应该按照下面的方式修改： ```python if not data_end.empty: # 后半段必须在年线以上运行（回踩年线） for index, row in data_end.iterrows(): if row['收盘'] < row['ma250']: return False if row['收盘'] < recent_lowest_row['收盘']: recent_lowest_row = row # date_diff = datetime.date(datetime.strptime(recent_lowest_row['日期'],...

@yhhhli Could you please help to answer the upper question?