[quant] support linear_relu_dynamic for qnnpack backend

[supriyar]

Stack from ghstack:

• [quant] Add support for linear_relu fusion for FP16 dynamic quant #63826
• [quant] Add op support for linear_relu_dynamic_fp16 #63824
• -> [quant] support linear_relu_dynamic for qnnpack backend #63820
• [quant][fx] Add support for dynamic linear + relu fusion (INT8) #63799

Summary:
Adds support in the operator directly to call relu operator if relu fusion is enabled.
Once QNNPACK natively supports relu fusion in the linear_dynamic this can be removed

Test Plan:
python test/test_quantization.py TestDynamicQuantizedLinear.test_qlinear

Reviewers:

Subscribers:

Tasks:

Tags:

Differential Revision: D30502813

[facebook-github-bot]

🔗 Helpful links

• 🧪 See artifacts and rendered test results at hud.pytorch.org/pr/63820
• 📄 Preview docs built from this PR

💊 CI failures summary and remediations

As of commit 1da6021 (more details on the Dr. CI page):

---

• 2/2 failures introduced in this PR

---

🕵️ 2 new failures recognized by patterns

The following CI failures do not appear to be due to upstream breakages:

pytorch_linux_bionic_cuda10_2_cudnn7_py3_9_gcc7_test1 (1/2)

Step: "Run tests" (full log | diagnosis details | 🔁 rerun)

Aug 26 00:22:05 ERROR [0.001s]: test_multitenancy (__main__.TCPStoreTest)

Aug 26 00:22:01   test_address_already_in_use (__main__.TCPStoreTest) ... ok (0.002s)
Aug 26 00:22:01   test_compare_set (__main__.TCPStoreTest) ... ok (0.002s)
Aug 26 00:22:02   test_init_pg_and_rpc_with_same_socket (__main__.TCPStoreTest) ... ok (0.071s)
Aug 26 00:22:03   test_multi_worker_with_fixed_world_size (__main__.TCPStoreTest) ... ok (0.005s)
Aug 26 00:22:03   test_multi_worker_with_nonfixed_world_size (__main__.TCPStoreTest) ... ok (0.006s)
Aug 26 00:22:03   test_multitenancy (__main__.TCPStoreTest) ... ERROR (0.001s)
Aug 26 00:22:05   test_numkeys_delkeys (__main__.TCPStoreTest) ... ok (2.013s)
Aug 26 00:22:05   test_set_get (__main__.TCPStoreTest) ... ok (0.003s)
Aug 26 00:22:05 
Aug 26 00:22:05 ======================================================================
Aug 26 00:22:05 ERROR [0.001s]: test_multitenancy (__main__.TCPStoreTest)
Aug 26 00:22:05 ----------------------------------------------------------------------
Aug 26 00:22:05 Traceback (most recent call last):
Aug 26 00:22:05   File "/var/lib/jenkins/workspace/test/distributed/test_store.py", line 179, in test_multitenancy
Aug 26 00:22:05     store1 = dist.TCPStore(addr, port, 1, True, multi_tenant=True)  # type: ignore[call-arg] # noqa: F841
Aug 26 00:22:05 RuntimeError: Address already in use
Aug 26 00:22:05 
Aug 26 00:22:05 ----------------------------------------------------------------------
Aug 26 00:22:05 Ran 19 tests in 3.985s
Aug 26 00:22:05 
Aug 26 00:22:05 FAILED (errors=1)

pytorch_linux_xenial_py3_clang7_onnx_ort_test2 (2/2)

Step: "Run tests" (full log | diagnosis details | 🔁 rerun)

Aug 25 23:57:43 AssertionError:

Aug 25 23:57:43   File "/opt/conda/lib/python3.6/site-packages/caffe2/python/operator_test/weighted_sample_test.py", line 18, in test_weighted_sample
Aug 25 23:57:43     batch=st.integers(min_value=0, max_value=128),
Aug 25 23:57:43   File "/opt/conda/lib/python3.6/site-packages/hypothesis/core.py", line 517, in test
Aug 25 23:57:43     result = self.test(*args, **kwargs)
Aug 25 23:57:43   File "/opt/conda/lib/python3.6/site-packages/caffe2/python/operator_test/weighted_sample_test.py", line 52, in test_weighted_sample
Aug 25 23:57:43     np.testing.assert_allclose(rand_indices[i], result_indices[i])
Aug 25 23:57:43   File "/opt/conda/lib/python3.6/site-packages/numpy/testing/_private/utils.py", line 1533, in assert_allclose
Aug 25 23:57:43     verbose=verbose, header=header, equal_nan=equal_nan)
Aug 25 23:57:43   File "/opt/conda/lib/python3.6/site-packages/numpy/testing/_private/utils.py", line 846, in assert_array_compare
Aug 25 23:57:43     raise AssertionError(msg)
Aug 25 23:57:43 AssertionError: 
Aug 25 23:57:43 Not equal to tolerance rtol=1e-07, atol=0
Aug 25 23:57:43 
Aug 25 23:57:43 Mismatched elements: 1 / 1 (100%)
Aug 25 23:57:43 Max absolute difference: 6.
Aug 25 23:57:43 Max relative difference: inf
Aug 25 23:57:43  x: array(6., dtype=float32)
Aug 25 23:57:43  y: array([0], dtype=int32)
Aug 25 23:57:43 
Aug 25 23:57:43 
Aug 25 23:57:43 You can reproduce this example by temporarily adding @reproduce_failure('3.44.6', b'APGMQbZP') as a decorator on your test case

---

This comment was automatically generated by Dr. CI (expand for details).

Follow this link to opt-out of these comments for your Pull Requests.

Please report bugs/suggestions to the (internal) Dr. CI Users group.

Click here to manually regenerate this comment.

[supriyar]

@supriyar has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

[supriyar]

@supriyar has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

[facebook-github-bot]

This pull request has been merged in 975f4cc.