Importance of opt_Level


#1

Hi.

I was running the tensorflow model with NNVM frontend.
and was getting sufficient output for all opt_levels.
But even after opt_level 3 i’m able to compile and build the outputs.

Is there any significance importance of opt_level after 3.
I’m getting results even at opt_level 5,10 and so on.
@srkreddy1238 @FrozenGene @tqchen
Any help here?

Thank you


#2

@Vinayak618 opt_level is mainly for enabling optimization passes like OpFusion, constant folding, …etc.
I think the max level now is 3 which includes all available passes.


#3

@srkreddy1238.

Am i doing anything wrong.
Will i get any error when i put opt_level more than 3.?
because while testing i have tried with opt_levels with 5, 10 and still able to get results for the same.

Thank you


#4

no problem with higher number.


#5

Thank you @srkreddy1238.

Also facing some unsupported operator error at opt_level=1 for both CUDA and OPENCL targets.

Below is the Error:

File "/home/ubuntu/.local/lib/python3.6/site-packages/topi-0.6.dev0-py3.6.egg/topi/cuda/reduction.py", line 132, in schedule_reduce
    traverse_after_reduce(outs[0].op)
  File "/home/ubuntu/.local/lib/python3.6/site-packages/topi-0.6.dev0-py3.6.egg/topi/cuda/reduction.py", line 115, in traverse_after_reduce
    traverse_after_reduce(tensor.op)
  File "/home/ubuntu/.local/lib/python3.6/site-packages/topi-0.6.dev0-py3.6.egg/topi/cuda/reduction.py", line 120, in traverse_after_reduce
    traverse_before_reduce(tensor.op)
  File "/home/ubuntu/.local/lib/python3.6/site-packages/topi-0.6.dev0-py3.6.egg/topi/cuda/reduction.py", line 103, in traverse_before_reduce
    traverse_before_reduce(tensor.op)
  File "/home/ubuntu/.local/lib/python3.6/site-packages/topi-0.6.dev0-py3.6.egg/topi/cuda/reduction.py", line 103, in traverse_before_reduce
    traverse_before_reduce(tensor.op)
  File "/home/ubuntu/.local/lib/python3.6/site-packages/topi-0.6.dev0-py3.6.egg/topi/cuda/reduction.py", line 103, in traverse_before_reduce
    traverse_before_reduce(tensor.op)
  [Previous line repeated 4 more times]
  File "/home/ubuntu/.local/lib/python3.6/site-packages/topi-0.6.dev0-py3.6.egg/topi/cuda/reduction.py", line 105, in traverse_before_reduce
    raise RuntimeError("Unsupported operator: %s" % operator.tag)
RuntimeError: Unsupported operator: 
Error during compile func

Any help here would be great.

Thank you


#6

As I remember there was similar issue in past and this is a limitation now.


#7

@srkreddy1238.
Limitation? So there is no way to resolve the issue?


#8

@eqy.
Any help here?


#9

Can you share more details about your issue (e.g., model, backend)?