apeforest commented on a change in pull request #15475: Add transpose_conv,
sorting and searching operator benchmarks to Opperf
URL: https://github.com/apache/incubator-mxnet/pull/15475#discussion_r303194813
##########
File path: benchmark/opperf/nd_operations/nn_conv_operators.py
##########
@@ -135,3 +135,50 @@ def run_convolution_operators_benchmarks(ctx=mx.cpu(),
dtype='float32', warmup=2
# Prepare combined results
mx_conv_op_results = merge_map_list(conv1d_benchmark_res +
conv2d_benchmark_res)
return mx_conv_op_results
+
+
+def run_transpose_convolution_operators_benchmarks(ctx=mx.cpu(),
dtype='float32', warmup=10, runs=50):
+ # Conv1DTranspose Benchmarks
+ conv1d_transpose_benchmark_res = []
+ for conv_data in [(32, 3, 256), (32, 3, 64)]:
+ conv1d_transpose_benchmark_res +=
run_performance_test([getattr(MX_OP_MODULE, "Deconvolution")],
+
run_backward=True,
+ dtype=dtype,
+ ctx=ctx,
+
inputs=[{"data": conv_data,
+
"weight": (3, 64, 3),
+
"bias": (64,),
+
"kernel": (3,),
+
"stride": (1,),
Review comment:
These parameters such as stride, dilate and pad seem to make impact to
performance. Would it make sense to test those variations?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
