rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument / FP16 performance on Volta
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-371654013
Both suggestions didn't help improve the speed unforunately. Using
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument / FP16 performance on Volta
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-372480640
Sorry I was digressing from the topic of the issue. Regarding the iterator
issue, we need
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument / FP16 performance on Volta
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-371654013
Both suggestions didn't help improve the speed unforunately. Using
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument / FP16 performance on Volta
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-371654013
Both suggestions didn't help improve the speed unforunately. Using
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument / FP16 performance on Volta
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-371654013
Both suggestions didn't help improve the speed unforunately. Using
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument / FP16 performance on Volta
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-371654013
Both suggestions didn't help improve the speed unforunately. Using
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument / FP16 performance on Volta
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-371654013
Both suggestions didn't help improve the speed unforunately. Using
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument / FP16 performance on Volta
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-367141388
Okay cool, I'll try to document that.
I was using
```
python
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument / FP16 performance on Volta
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-367141388
Okay cool, I'll try to document that.
I was using
```
python
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument / FP16 performance on Volta
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-367141388
Okay cool, I'll try to document that.
I was using ```
python train_imagenet.py
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument / FP16 performance on Volta
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-367142407
I have cudnn v7005 and cuda 9.0
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument / FP16 performance on Volta
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-367141388
Okay cool, I'll try to document that.
I was using the maximum batch size which is
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument / FP16 performance on Volta
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-367141388
Okay cool, I'll try to document that.
I was using the maximum batch size which is
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument / FP16 performance on Volta
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-367124681
I'm running this command.
`python train_cifar10.py --batch-size 256 --network
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument / FP16 performance on Volta
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-367124681
I'm running this command.
`python train_cifar10.py --batch-size 256 --network
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-366835890
But for Resnet 110 on Cifar10, fp16 is much slower. Do you see something
fishy here? There are barely any operations
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-366823518
Hey @KellenSunderland
I ran Resnet50 with Imagenet and got about 70% speedup. Some of the top ones
don't seem
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-366823518
Hey @KellenSunderland
These are the top 30 activities on the GPU. I'm running Resnet 50 on
imagenet data with
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-366788577
Thanks @KellenSunderland . I am unable to run nvprof . Looks like this is a
known issue
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-366788577
Thanks @KellenSunderland . I am unable to run nvprof . Looks like this is a
known issue
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-366788577
Thanks @KellenSunderland . I am unable to run nvprof . Looks like this is a
known issue
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-365780748
Thanks, this
[blogpost](http://docs.nvidia.com/deeplearning/sdk/mixed-precision-training/index.html)
mentions
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-365780748
Thanks, this
[blogpost](http://docs.nvidia.com/deeplearning/sdk/mixed-precision-training/index.html)
mentions
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-365769322
Thanks for the explanation.
Btw, is training in fp16 supposed to be ~2x faster than fp32 for a given
batch
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-365769322
Thanks for the explanation.
Btw, is training in fp16 supposed to be ~2x faster than fp32 for a given
batch
rahul003 commented on issue #9774: mx.io.ImageRecordIter does not respect dtype
argument
URL:
https://github.com/apache/incubator-mxnet/issues/9774#issuecomment-365769322
Thanks for the explanation.
Btw, is training in fp16 supposed to be ~2x faster than fp32 for a given
batch
26 matches
Mail list logo
