These are chat archives for dmlc/mxnet

Mar 2018
Mar 16 2018 11:10
Hi, I'm trying to compile a mxnet model using NNVM as described here: Everything works fine, except when I'm trying to compile with cuda as target with batchsize > 1. All NNVM tutorials and examples I could find also only use batchsize 1. The error is: RuntimeError: Batch size: 32 is too large for this schedule (topi/cuda/", line 527, in schedule_conv2d_nchw)