[quant] Support 2 dim input in quantized batchnorm 1d

[jerryzh168]

Stack from ghstack:

• [quant] Support 2 dim input in quantized batchnorm 1d #51597 [quant] Support 2 dim input in quantized batchnorm 1d

Summary:
aliging quantized batchnorm behavior with fp batchnorm

Test Plan:
python test/test_quantization.py TestQuantizedOps.test_batch_norm
python test/test_quantization.py TestQuantizedOps.test_batch_norm_relu

Reviewers:

Subscribers:

Tasks:

Tags:

Differential Revision: D26212489

[facebook-github-bot]

💊 CI failures summary and remediations

As of commit 0309da1 (more details on the Dr. CI page):

---

• 2/2 failures possibly* introduced in this PR
  • 1/2 non-CircleCI failure(s)

---

🕵️ 1 new failure recognized by patterns

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

pytorch_linux_xenial_py3_6_gcc5_4_test (1/1)

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

Feb 03 01:02:41 [E request_callback_no_python.cpp:653] Received error while processing request type 258: RuntimeError: Can not pickle torch.futures.Future

Feb 03 01:02:41 At:
Feb 03 01:02:41   /opt/conda/lib/python3.6/site-packages/torch/distributed/rpc/internal.py(120): serialize
Feb 03 01:02:41   /opt/conda/lib/python3.6/site-packages/torch/distributed/rpc/internal.py(172): serialize
Feb 03 01:02:41 
Feb 03 01:02:41 [E request_callback_no_python.cpp:653] Received error while processing request type 258: RuntimeError: Can not pickle torch.futures.Future
Feb 03 01:02:41 
Feb 03 01:02:41 At:
Feb 03 01:02:41   /opt/conda/lib/python3.6/site-packages/torch/distributed/rpc/internal.py(120): serialize
Feb 03 01:02:41   /opt/conda/lib/python3.6/site-packages/torch/distributed/rpc/internal.py(172): serialize
Feb 03 01:02:41 
Feb 03 01:02:41 [E request_callback_no_python.cpp:653] Received error while processing request type 258: RuntimeError: Can not pickle torch.futures.Future
Feb 03 01:02:41 
Feb 03 01:02:41 At:
Feb 03 01:02:41   /opt/conda/lib/python3.6/site-packages/torch/distributed/rpc/internal.py(120): serialize
Feb 03 01:02:41   /opt/conda/lib/python3.6/site-packages/torch/distributed/rpc/internal.py(172): serialize
Feb 03 01:02:41 
Feb 03 01:02:41 ok (1.942s)
Feb 03 01:02:43   test_return_future_remote (__main__.ProcessGroupRpcTestWithSpawn) ... RPC was initialized with the PROCESS_GROUP backend which is deprecated and slated to be removed and superseded by the TENSORPIPE backend. It is recommended to migrate to the TENSORPIPE backend.
Feb 03 01:02:43 RPC was initialized with the PROCESS_GROUP backend which is deprecated and slated to be removed and superseded by the TENSORPIPE backend. It is recommended to migrate to the TENSORPIPE backend.
Feb 03 01:02:43 RPC was initialized with the PROCESS_GROUP backend which is deprecated and slated to be removed and superseded by the TENSORPIPE backend. It is recommended to migrate to the TENSORPIPE backend.
Feb 03 01:02:43 RPC was initialized with the PROCESS_GROUP backend which is deprecated and slated to be removed and superseded by the TENSORPIPE backend. It is recommended to migrate to the TENSORPIPE backend.

---

ci.pytorch.org: 1 failed

• Failed: pr/pytorch-linux-bionic-rocm3.10-py3.6

---

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.

[vkuzo]

looks good, thank you! Did we check that this is the same behavior as fp32 for dim==2? I don't think that needs to be in the unit tests, just confirming as I haven't actually looked into the fp32 op implementation for this issue.

[jerryzh168]

looks good, thank you! Did we check that this is the same behavior as fp32 for dim==2? I don't think that needs to be in the unit tests, just confirming as I haven't actually looked into the fp32 op implementation for this issue.

yeah we have tests for both quantized batchnorm and batchnorm_relu that matches the reference implementation with dequant - float batchnorm - quant

[facebook-github-bot]

This pull request has been merged in d8742ee.