Attention_ocr

Attention_ocr copied to clipboard

hi, i tryed to run your code and got some cuda errors. could you provide witch versions of pytorch/torchvision and cuda where you using. thanks

malfonso0

lt573152873

>Traceback (most recent call last): > File "train.py", line 58, in print_interval=opt.print_interval, eval_step=opt.eval_step, save_step=opt.save_step, save_dir=opt.save_dir, use_gpu=opt.use_gpu)` File "C:\Users\vzeml\Desktop\w\w\ocr\core\train_engine.py", line 48, in fit for i, data in enumerate(train_data): File "C:\Users\vzeml\Anaconda3\lib\site-packages\torch\utils\data\dataloader.py", line...

rainsummer613

不定长输出

1

你好，修改了maxlength 和 DECODER_OUTPUT_FRAME 最终的输出长度还是 和 DECODER_OUTPUT_FRAME 一样，如何输出 不定长结果呢？

tomjerrygithub

@JimmyHHua 你的attention里面怎么那么多全连接层， 这样计算量不会很大吗，训练起来也会很慢的吧

sporterman

不是10个字的能识别对吗？

5

dlml