Performance


Topic Replies Activity
Does lazy evaluation make batch size less important? 2 February 12, 2019
Scalable (parallel) recordio files creation? 4 February 8, 2019
Append records to recordio dataset? 3 February 7, 2019
How to make MxNet use only specific CPU cores 4 January 30, 2019
Overlap gradient communication with backward pass 3 January 16, 2019
Why CPU load is so heavy during training? How can I reduce it 5 January 16, 2019
MxNet C Api Executor Reshape Question 6 January 14, 2019
Hybrid training speed is 20% slower than pytorch 6 January 11, 2019
Color Blind SSD (VGG-16) model 2 January 7, 2019
`MXImperativeInvokeEx` is taking a long time 9 January 6, 2019
Multiple dataloader will slow the training performance 2 December 27, 2018
NDArray "cold start" on GPU? 3 December 21, 2018
Slow speed of mallocing gpu memory using mxnet built from source 2 December 10, 2018
MXNet crashing, likely memory corruption 10 January 10, 2018
Is mx.nd.array thread safety? 2 November 26, 2018
Understanding MXNet multi-gpu performance 8 November 5, 2018
When to set CUDNN_AUTOTUNE_DEFAULT to 0? 2 October 23, 2018
Possible Memory Leak 2 October 22, 2018
Efficiently accessing arbitrary NDArray batches in mxnet 4 October 22, 2018
Guidance for big data loading with MXNet 2 October 17, 2018
GPU memory in training vs bind/load 3 October 12, 2018
Accuracy Drop when Inference in the autograd.record scope 2 October 7, 2018
How to use all available cores in mxnet 2 September 28, 2018
Mxnet GPU freezes python 4 September 26, 2018
MXNet Distributed Training - Meetup in Palo Alto 10/9 2 September 19, 2018
Memory issue with Module forward function 3 September 19, 2018
Speed with and without evaluation metric of the model 6 August 30, 2018
MXnet error: Backward Norm Operator not implemented in GPU 5 August 24, 2018
Difference between nightly build and 1.2 version? 3 August 20, 2018
Gluon implementation much slower than Symbolic 10 August 20, 2018