You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@mxnet.apache.org by GitBox <gi...@apache.org> on 2018/11/18 03:20:35 UTC
[GitHub] connermcb opened a new issue #13314: Check failure during
exec$update.arg.arrays - dims mismatch
connermcb opened a new issue #13314: Check failure during exec$update.arg.arrays - dims mismatch
URL: https://github.com/apache/incubator-mxnet/issues/13314
## Description
bucket.plan indices appear not to be updating correctly causing attribute check to fail.
## Environment info (Required)
R version 3.5.1 (2018-07-02)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows >= 8 x64 (build 9200)
Matrix products: default
locale:
[1] LC_COLLATE=English_United States.1252 LC_CTYPE=English_United States.1252
[3] LC_MONETARY=English_United States.1252 LC_NUMERIC=C
[5] LC_TIME=English_United States.1252
attached base packages:
[1] stats graphics grDevices utils datasets methods base
other attached packages:
[1] bindrcpp_0.2.2 mxnet_1.3.0 stringi_1.2.4 stringr_1.3.1 readr_1.1.1
loaded via a namespace (and not attached):
[1] Rcpp_0.12.17 plyr_1.8.4 pillar_1.3.0 compiler_3.5.1
[5] RColorBrewer_1.1-2 influenceR_0.1.0 bindr_0.1.1 viridis_0.5.1
[9] tools_3.5.1 digest_0.6.15 jsonlite_1.5 viridisLite_0.3.0
[13] gtable_0.2.0 tibble_1.4.2 rgexf_0.15.3 pkgconfig_2.0.1
[17] rlang_0.2.1 igraph_1.2.1 rstudioapi_0.7 yaml_2.1.19
[21] gridExtra_2.3 downloader_0.4 DiagrammeR_1.0.0 dplyr_0.7.6
[25] htmlwidgets_1.2 hms_0.4.2 grid_3.5.1 tidyselect_0.2.4
[29] glue_1.2.0 R6_2.2.2 tcltk_3.5.1 Rook_1.1-1
[33] XML_3.98-1.12 ggplot2_3.0.0 purrr_0.2.5 tidyr_0.8.1
[37] magrittr_1.5 codetools_0.2-15 scales_1.0.0 htmltools_0.3.6
[41] assertthat_0.2.0 colorspace_1.3-2 brew_1.0-6 visNetwork_2.0.4
[45] lazyeval_0.2.1 munsell_0.5.0 crayon_1.3.4
## Error Message:
Error in exec$update.arg.arrays(arg.arrays, match.name, skip.null) :
[19:39:48] c:\jenkins\workspace\mxnet\mxnet\src\operator\tensor\../elemwise_op_common.h:123: Check failed: assign(&dattr, (*vec)[i]) Incompatible attr in node at 0-th output: expected [32,10], got [32,42]
## Minimum reproducible example
https://github.com/connermcb/next_word_predictor
## Steps to reproduce
Just run small_rnn.R.
## What have you tried to solve it?
The error is thrown by a header file, and I'm not able to troubleshoot C++ effectively. Interestingly, when restarting the R session the model makes it to the second batch before throwing the error:
`expected [32,42], got [32,10] # 42 and 10 are the bucket.size minus 1 for the first two buckets in plan`
```
> train_data$batch
[1] 2
> head(train_data$bucket.plan)
11 43 8 6 7 3
1 1 1 1 1 1
> train_data$bucketID
43
1
```
It looks like the check looks for the dimensions of the second bucket in plan but still finds the first.
But if, I then try running only mx.model.buckets again:
`[32,10], got [32,7]`
```
> train_data$batch
[1] 1
> head(train_data$bucket.plan)
11 43 8 6 7 3
1 1 1 1 1 1
> train_data$bucketID
11
1
```
It appear that the check is now looking for the dimensions of the first bucket in the plan, but getting the dimensions of the third, as though there were a index value that's not updating or resetting correctly.
If I run only the mx.model.buckets function one or more times from there without resetting anything:
`[32,10], got [32,42]`
```
> train_data$batch
[1] 1
> head(train_data$bucket.plan)
11 43 8 6 7 3
1 1 1 1 1 1
> train_data$bucketID
11
1
```
It appears to be looking for dims of first bucket but getting those of the second.
Is this issue related to how I have the model set up or an issue with the code in elemwise_op_common.h ?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services