Ricahrd-Li opened a new issue #17663: [Question About Profiler] Is there a way 
to know which number of layer of neural network the tensor belongs to in the 
result of MXNet profiller? 
URL: https://github.com/apache/incubator-mxnet/issues/17663
 
 
   ## Description
   (A clear and concise description of what the feature is.)
   - If the proposal is about a new model, provide description of what the 
model is.
   - If the proposal is about an API, provide mock examples if possible.
   
   Hi there! 
   Because of need in work, I am wondering is there a way to know **which 
number of layer of the neural network** the tensor belongs to in the result of 
MXNet profiller? 
   
   For example,
   In the result of profiler, I see result like this:
   ```
      {
           "name": "Convolution",
           "cat": "operator",
           "ph": "B",
           "ts": 1581521825791464,
           "pid": 16,
           "tid": 3803165864448931469
       }
   ,
       {
           "name": "Convolution",
           "cat": "operator",
           "ph": "E",
           "ts": 1581521825791916,
           "pid": 16,
           "tid": 3803165864448931469
       }
   ,
   
       {
           "name": "BatchNorm",
           "cat": "operator",
           "ph": "B",
           "ts": 1581521825791920,
           "pid": 16,
           "tid": 3803165864448931469
       }
   ```
   I want to specify the number of layer in the profiling result, like:
   ```
      {
           "name": "Convolution_2",
           "cat": "operator",
           "ph": "B",
           "ts": 1581521825791464,
           "pid": 16,
           "tid": 3803165864448931469
       }
   ,
       {
           "name": "Convolution_2",
           "cat": "operator",
           "ph": "E",
           "ts": 1581521825791916,
           "pid": 16,
           "tid": 3803165864448931469
       }
   ,
   
       {
           "name": "BatchNorm_3",
           "cat": "operator",
           "ph": "B",
           "ts": 1581521825791920,
           "pid": 16,
           "tid": 3803165864448931469
       }
   ```
   From the result above, I will know they are the second convolution layer and 
third batchnorm layer in my DNN (which may have 10 convolution layers and 5 
batchnorm layers with different tensor shape and forward & backword computation 
time)
   
   In the profiler of Horovod (called **Horovod timeline**), I am able to 
modify the source code and get "tensor_name" printed in the profiling result, 
which contains the number of layer. But in MXNet, I am still wondering how to 
achieve this. 
   
   Thanks!
   
   ## References
   - list reference and related literature
   - list known implementations
   

----------------------------------------------------------------
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

Reply via email to