Open garfieldnate opened 7 years ago
can you give me the terminal trace, of what it is outputting. This is strange. Also, I don't remember exactly, as this project is a year old. However, usually changing the memory size etc should help.
Thanks for looking at it. I appreciate your time.
Another one of the scripts failed, too, but with different error messages. I'm attaching output corresponding to the names of the scripts I ran:
I have the same issue, the generated summary is empty. Is there any other step before running train_script_gru_simple_no_attn.py ? I have attached a snippet of my screen.
Thanks.
Were you getting some error and for that, you commented some portion of it ? Then it could be possible to generate empty summary.
No, actually I just followed the steps from your readme file, with the same version of tensorflow installed as written in the requirements file. In the first few iterations, some generated words are printed on screen and after Step No. 8 or 9 generated summary becomes empty.
Same thing happens for me. After a few steps all the summaries get empty for the no_attn scripts. Also the stacked simple attn gets an error (Invalid argument: cannot parse tensor from proto: dtype: DT_FLOAT). I was wondering if you found a fix for this?
Did anyone find a solution for it?
It's generating empty summaries in my case as well and I haven't made any changes in the code before running it.
Same empty issue!
So any solution? Same issue as well. I'm in step 38. Don't know if i should stop because it's not generating.
I'm still finding solution,but the result is nagative....
Is it possible because of memory issue? How much memory is it using?
Maybe not ,cause when I use GruSimplieNoattention model there isn't OOM error occured.
And now I transfer to textsum model,but the output is still not ideal.
I let
train_script_gru_simple_no_attn.py
run for a long time, and each iteration the generated summary is completely empty. Is this normal? How many iterations is it supposed to run before exiting? After 1500 iterations the generated summaries are still empty.