access2rohit commented on issue #17331:
URL:
https://github.com/apache/incubator-mxnet/issues/17331#issuecomment-656790489
> upon investigation openBLAS needs to be built with specific flag to
support int64_t signatures and MKL will support long long int signatures(in
which case
access2rohit commented on issue #17331:
URL:
https://github.com/apache/incubator-mxnet/issues/17331#issuecomment-656772572
> Has the large tensor for numpy array been supported?
We looked at numpy files inside MXNet and they are using index_t for
iterating over elements in
access2rohit commented on issue #17331:
URL:
https://github.com/apache/incubator-mxnet/issues/17331#issuecomment-656769897
> Make openBLAS compatible with Large Tensor support and merge the PR for
Enabling Large Tensor Support so that default PyPi users of MXNet can already
benefit from
access2rohit commented on issue #17331:
URL:
https://github.com/apache/incubator-mxnet/issues/17331#issuecomment-656499453
Currently Large Tensor Support work on all operators implemented in MXNet
and MKLDNN also supports int64. CUDA kernels written inside MXNET both
generic(cpu/gpu) and
access2rohit commented on issue #17331:
URL:
https://github.com/apache/incubator-mxnet/issues/17331#issuecomment-650509402
@apeforest @sandeep-krishnamurthy @szha
PR's to enable Large Tensor Support as default in master are divided into
two stages:
Stage1: Unix CPU/GPU and
access2rohit commented on issue #17331:
URL:
https://github.com/apache/incubator-mxnet/issues/17331#issuecomment-650509034
PR: https://github.com/apache/incubator-mxnet/pull/17882 fixes regression in
SSD. Following are the new results for SSD run:
Code | SSD 1 Epoch time (sec) |
access2rohit commented on issue #17331:
URL:
https://github.com/apache/incubator-mxnet/issues/17331#issuecomment-625542495
@apeforest THe profiling done by @JonTanS was done long back using
mxnet-1.6in november. These results are using current master branch of MXNet,
bert scripts have
access2rohit commented on issue #17331:
URL:
https://github.com/apache/incubator-mxnet/issues/17331#issuecomment-622665471
@szha @eric-haibin-lin @apeforest
With current master and new broadcast_axis changes on p3.16xl single GPU
training run.
Bert Run Command:
```
access2rohit commented on issue #17331:
URL:
https://github.com/apache/incubator-mxnet/issues/17331#issuecomment-622660352
[new_bert_train.log](https://github.com/apache/incubator-mxnet/files/4567005/new_bert_train.log)