You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by GitBox <gi...@apache.org> on 2022/06/15 07:50:04 UTC

[GitHub] [tvm] yuanfz98 opened a new pull request, #11729: [Pytorch] Add quantized::leaky_relu

yuanfz98 opened a new pull request, #11729:
URL: https://github.com/apache/tvm/pull/11729

   This PR implements quantized::leaky_relu. Related issue: https://github.com/apache/tvm/issues/11451.


-- 
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.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


[GitHub] [tvm] yuanfz98 commented on a diff in pull request #11729: [Pytorch] Add quantized::leaky_relu

Posted by GitBox <gi...@apache.org>.
yuanfz98 commented on code in PR #11729:
URL: https://github.com/apache/tvm/pull/11729#discussion_r898043135


##########
tests/python/frontend/pytorch/qnn_test.py:
##########
@@ -347,6 +360,7 @@ def test_quantized_modules():
         # sample outputs
         """
         relu 0.0039215684 2.6052087e-08 0.9999933567176871
+        leaky_relu 0.25097468 0.16687508 0.0036936649659863948

Review Comment:
   Oups! Sorry for the mistake !
   ```
   dequantized = dequant * _op.nn.leaky_relu(dequant, alpha)
   ```
   ->
   ```
   dequantized = _op.nn.leaky_relu(dequant, alpha)
   ```
   
   ```
   tvm_result= [[[[0.42745024 0.7333321  0.9333317  ... 0.18039185 0.4627443
       0.06666655]
      [0.5607833  0.65490085 0.1333331  ... 0.19607809 0.1333331
       0.42352867]
      [0.84705734 0.83921427 0.07058811 ... 0.47843054 0.25490153
       0.5215677 ]
      ...
      [0.9568611  0.34117588 0.07450967 ... 0.10196061 0.8078417
       0.3607837 ]
      [0.4627443  0.24705839 0.6156852  ... 0.47843054 0.8980377
       0.18039185]
      [0.8941161  0.9686258  0.19215654 ... 0.0823528  0.6784302
       0.27843088]]
   
     [[0.69803804 0.7490183  0.14509779 ... 0.06274499 0.01960781
       0.28235245]
      [0.21960747 0.1137253  0.32548964 ... 0.7647046  0.75686145
       0.2156859 ]
      [0.97254735 0.98823357 0.06666655 ... 0.79999864 0.0509803
       0.45882273]
      ...
      [0.70195955 0.8313711  0.40784243 ... 0.06666655 0.96078265
       0.662744  ]
      [0.74509674 0.52941084 0.76862615 ... 0.09411748 0.5019599
       0.21176434]
      [0.14117622 0.6784302  0.74509674 ... 0.88627297 0.07843123
       0.56862646]]
   
     [[0.32156807 0.30588183 0.01960781 ... 0.70195955 0.7411752
       0.09019592]
      [0.68235177 0.6901949  0.33725432 ... 0.12156842 0.52941084
       0.9686258 ]
      [0.28627402 0.3098034  0.7921555  ... 0.43921494 0.98431206
       0.01568625]
      ...
      [0.9764689  0.8352927  0.49019524 ... 0.7098027  0.6470577
       0.8941161 ]
      [0.09411748 0.31372494 0.921567   ... 0.54117554 0.09019592
       0.00392156]
      [0.8196064  0.49019524 0.1333331  ... 0.05490187 0.02745093
       0.2588231 ]]]]
   pt_result= [[[[0.42745024 0.7333321  0.9333317  ... 0.18039185 0.4627443
       0.06666655]
      [0.5607833  0.65490085 0.1333331  ... 0.19607809 0.1333331
       0.42352867]
      [0.84705734 0.83921427 0.07058811 ... 0.47843054 0.25490153
       0.5215677 ]
      ...
      [0.9568611  0.34117588 0.07450967 ... 0.10196061 0.8078417
       0.3607837 ]
      [0.4627443  0.24705839 0.6156852  ... 0.47843054 0.8980377
       0.18039185]
      [0.8941161  0.9686258  0.19215654 ... 0.0823528  0.6784302
       0.27843088]]
   
     [[0.69803804 0.7490183  0.14509779 ... 0.06274499 0.01960781
       0.28235245]
      [0.21960747 0.1137253  0.32548964 ... 0.7647046  0.75686145
       0.2156859 ]
      [0.97254735 0.98823357 0.06666655 ... 0.79999864 0.0509803
       0.45882273]
      ...
      [0.70195955 0.8313711  0.40784243 ... 0.06666655 0.96078265
       0.662744  ]
      [0.74509674 0.52941084 0.76862615 ... 0.09411748 0.5019599
       0.21176434]
      [0.14117622 0.6784302  0.74509674 ... 0.88627297 0.07843123
       0.56862646]]
   
     [[0.32156807 0.30588183 0.01960781 ... 0.70195955 0.7411752
       0.09019592]
      [0.68235177 0.6901949  0.33725432 ... 0.12156842 0.52941084
       0.9686258 ]
      [0.28627402 0.3098034  0.7921555  ... 0.43921494 0.98431206
       0.01568625]
      ...
      [0.9764689  0.8352927  0.49019524 ... 0.7098027  0.6470577
       0.8941161 ]
      [0.09411748 0.31372494 0.921567   ... 0.54117554 0.09019592
       0.00392156]
      [0.8196064  0.49019524 0.1333331  ... 0.05490187 0.02745093
       0.2588231 ]]]]
   ```



-- 
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.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


[GitHub] [tvm] masahi commented on a diff in pull request #11729: [Pytorch] Add quantized::leaky_relu

Posted by GitBox <gi...@apache.org>.
masahi commented on code in PR #11729:
URL: https://github.com/apache/tvm/pull/11729#discussion_r897824739


##########
tests/python/frontend/pytorch/qnn_test.py:
##########
@@ -347,6 +360,7 @@ def test_quantized_modules():
         # sample outputs
         """
         relu 0.0039215684 2.6052087e-08 0.9999933567176871
+        leaky_relu 0.25097468 0.16687508 0.0036936649659863948

Review Comment:
   Isn't the result too off from PyTorch?



-- 
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.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


[GitHub] [tvm] masahi merged pull request #11729: [Pytorch] Add quantized::leaky_relu

Posted by GitBox <gi...@apache.org>.
masahi merged PR #11729:
URL: https://github.com/apache/tvm/pull/11729


-- 
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.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org