MasterJH5574 opened a new issue, #14760:
URL: https://github.com/apache/tvm/issues/14760
Hello community members! Today when I running all the unit tests under
`tests/python/relax`, I found the tests under `test_relay_translator.py`
consumes quite much time. They greatly increases the time needed to a round of
testing.
```
>>> pytest -vv test_relay_translator.py --durations=0
rootdir: /home/ruihangl/tvm/tests/python/relax
collected 22 items
test_relay_translator.py::test_verify_e2e_translation_cpu[NCHW-1-image_shape0]
SKIPPED (take too much time)
[ 4%]
test_relay_translator.py::test_verify_e2e_translation_cpu[NHWC-1-image_shape1]
SKIPPED (take too much time)
[ 9%]
test_relay_translator.py::test_verify_e2e_translation_gpu[NCHW-1-image_shape0]
SKIPPED (take too much time)
[ 13%]
test_relay_translator.py::test_verify_e2e_translation_gpu[NHWC-1-image_shape1]
SKIPPED (take too much time)
[ 18%]
test_relay_translator.py::test_verify_extracted_tasks_cpu[NCHW-1-image_shape0-structural]
PASSED
[ 22%]
test_relay_translator.py::test_verify_extracted_tasks_cpu[NCHW-1-image_shape0-ignore-ndarray]
PASSED
[ 27%]
test_relay_translator.py::test_verify_extracted_tasks_cpu[NCHW-1-image_shape0-anchor-block]
PASSED
[ 31%]
test_relay_translator.py::test_verify_extracted_tasks_cpu[NHWC-1-image_shape1-structural]
PASSED
[ 36%]
test_relay_translator.py::test_verify_extracted_tasks_cpu[NHWC-1-image_shape1-ignore-ndarray]
PASSED
[ 40%]
test_relay_translator.py::test_verify_extracted_tasks_cpu[NHWC-1-image_shape1-anchor-block]
PASSED
[ 45%]
test_relay_translator.py::test_verify_extracted_tasks_gpu[NCHW-1-image_shape0-structural]
PASSED
[ 50%]
test_relay_translator.py::test_verify_extracted_tasks_gpu[NCHW-1-image_shape0-ignore-ndarray]
PASSED
[ 54%]
test_relay_translator.py::test_verify_extracted_tasks_gpu[NCHW-1-image_shape0-anchor-block]
PASSED
[ 59%]
test_relay_translator.py::test_verify_extracted_tasks_gpu[NHWC-1-image_shape1-structural]
PASSED
[ 63%]
test_relay_translator.py::test_verify_extracted_tasks_gpu[NHWC-1-image_shape1-ignore-ndarray]
PASSED
[ 68%]
test_relay_translator.py::test_verify_extracted_tasks_gpu[NHWC-1-image_shape1-anchor-block]
PASSED
[ 72%]
test_relay_translator.py::test_single_dynamic_dim PASSED
[ 77%]
test_relay_translator.py::test_multiple_dynamic_dims PASSED
[ 81%]
test_relay_translator.py::test_layout_transform PASSED
[ 86%]
test_relay_translator.py::test_translate_op_with_tir PASSED
[ 90%]
test_relay_translator.py::test_translate_tuple_arg PASSED
[ 95%]
test_relay_translator.py::test_append_op_attrs PASSED
[100%]
===================================================================================
slowest durations
====================================================================================
33.96s call
test_relay_translator.py::test_verify_extracted_tasks_gpu[NCHW-1-image_shape0-anchor-block]
33.80s call
test_relay_translator.py::test_verify_extracted_tasks_gpu[NCHW-1-image_shape0-ignore-ndarray]
33.70s call
test_relay_translator.py::test_verify_extracted_tasks_gpu[NCHW-1-image_shape0-structural]
29.91s call
test_relay_translator.py::test_verify_extracted_tasks_cpu[NCHW-1-image_shape0-anchor-block]
29.76s call
test_relay_translator.py::test_verify_extracted_tasks_cpu[NCHW-1-image_shape0-ignore-ndarray]
29.58s call
test_relay_translator.py::test_verify_extracted_tasks_cpu[NCHW-1-image_shape0-structural]
28.17s call
test_relay_translator.py::test_verify_extracted_tasks_cpu[NHWC-1-image_shape1-structural]
28.16s call
test_relay_translator.py::test_verify_extracted_tasks_gpu[NHWC-1-image_shape1-anchor-block]
28.04s call
test_relay_translator.py::test_verify_extracted_tasks_cpu[NHWC-1-image_shape1-anchor-block]
28.02s call
test_relay_translator.py::test_verify_extracted_tasks_gpu[NHWC-1-image_shape1-ignore-ndarray]
28.00s call
test_relay_translator.py::test_verify_extracted_tasks_gpu[NHWC-1-image_shape1-structural]
27.92s call
test_relay_translator.py::test_verify_extracted_tasks_cpu[NHWC-1-image_shape1-ignore-ndarray]
0.34s call test_relay_translator.py::test_single_dynamic_dim
0.30s call test_relay_translator.py::test_multiple_dynamic_dims
0.22s call test_relay_translator.py::test_translate_op_with_tir
...
```
The tests are run on our research group server with CPU AMD 3970X and GPU
RTX 4090. I imagine the tests will take no less time when run on other machines.
So here I want to propose to skip these tests. Actually these tests might be
far from being called “unit.” We perhaps want to separate them into another
directory.
--
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]
