denise-k opened a new issue #8610:
URL: https://github.com/apache/tvm/issues/8610


   # Context
   
   CI is taking a while (see #8552 and 
[build#1384](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/)), and I 
am tracking short-term fixes to alleviate this
   
   # Summary of the issue
   
   Each of the frontends in TVM is testing multiple common networks (e.g. 
`resnet_50, inception_v1, googlenet`). Functionally, the networks are the same 
once they get lowered into TVM, which means that there is a lot of duplicate 
coverage for each frontend.
   
   Each network test may only take seconds or minutes, but due to the 
ever-growing and overlapping `number of networks tested * number of frameworks 
supported`, this is now taking up over an hour of CI time.
   
   Here is a list of these network tests:
   
   | Frontend | Model | Runtime |
   | ---- | ---- | ----|
   | Tensorflow | 
[ssd](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/cython.tests.python.frontend.tensorflow/test_forward/Integration_Test___frontend__GPU___test_forward_ssd)
 | 11 min |
   | TensorFlow | [resnet 
v2](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/cython.tests.python.frontend.tensorflow/test_forward/Integration_Test___frontend__GPU___test_forward_resnetv2)
       | 2 min 30 sec |
   | TensorFlow | 
[ptb](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/cython.tests.python.frontend.tensorflow/test_forward/Integration_Test___frontend__GPU___test_forward_ptb)
        | 2 min 43 sec |
   | TensorFlow | 
[mobilenet](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/cython.tests.python.frontend.tensorflow/test_forward/Integration_Test___frontend__GPU___test_forward_mobilenet)
 | 43 sec |
   | PyTorch | [densenet 
121](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/cython.tests.python.frontend.pytorch/test_forward/Integration_Test___frontend__GPU___test_densenet121)
 | 2 min 32 sec |
   | PyTorch | 
[googlenet](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/cython.tests.python.frontend.pytorch/test_forward/Integration_Test___frontend__GPU___test_googlenet)
 | 1 min 6 sec |
   | PyTorch | [inception 
v3](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/cython.tests.python.frontend.pytorch/test_forward/Integration_Test___frontend__GPU___test_inception_v3)
 | 1 min 34 sec |
   | PyTorch | [mnasnet 
0.5](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/cython.tests.python.frontend.pytorch/test_forward/Integration_Test___frontend__GPU___test_mnasnet0_5)
 | 53 sec |
   | PyTorch | [mobilenet 
v2](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/cython.tests.python.frontend.pytorch/test_forward/Integration_Test___frontend__GPU___test_mobilenet_v2)
 | 49 sec |
   | PyTorch | 
[resnet18](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/cython.tests.python.frontend.pytorch/test_forward/Integration_Test___frontend__GPU___test_resnet18)
 | 30 sec |
   | PyTorch | 
[segmentation](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/cython.tests.python.frontend.pytorch/test_forward/Integration_Test___frontend__GPU___test_segmentation_models)
 | 3 min 52 sec |
   | PyTorch | [squeezenet 
1.0](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/cython.tests.python.frontend.pytorch/test_forward/Integration_Test___frontend__GPU___test_squeezenet1_0)
 | 17 sec |
   | PyTorch | [squeezenet 
1.1](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/cython.tests.python.frontend.pytorch/test_forward/Integration_Test___frontend__GPU___test_squeezenet1_1)
 | 13 sec |
   | Caffe | [inception 
v1](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/cython.tests.python.frontend.caffe/test_forward/Integration_Test___frontend__CPU___test_forward_Inceptionv1)
 | 31 sec |
   | Caffe | [mobilenet 
v2](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/cython.tests.python.frontend.caffe/test_forward/Integration_Test___frontend__CPU___test_forward_Mobilenetv2)
 | 31 sec |
   | Caffe | [resnet 
50](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/cython.tests.python.frontend.caffe/test_forward/Integration_Test___frontend__CPU___test_forward_Resnet50)
 | 44 sec |
   | ONNX | 
[densenet](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/cython.tests.python.frontend.onnx/test_forward/Integration_Test___frontend__GPU___test_densenet)
 | 2 min 2 sec |
   | CoreML | 
[mobilenet](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/cython.tests.python.frontend.coreml/test_forward/Integration_Test___frontend__GPU___test_mobilenet_checkonly)
 | 33 sec |
   | CoreML | 
[resnet](https://ci.tlcpack.ai/job/tvm/job/main/1384/testReport/cython.tests.python.frontend.coreml/test_forward/Integration_Test___frontend__GPU___test_resnet50_checkonly)
 | 1 min 12 sec | 
   
   
   # Suggested actions
   * Separate out full network test cases from frontend tests
   * Make sure critical networks are tested in integration tests
   * Move full-network frontend tests to a nightly or non-PR-gating CI pipeline
   


-- 
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: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


Reply via email to