-
Notifications
You must be signed in to change notification settings - Fork 25.6k
Closed
Labels
module: flaky-testsProblem is a flaky test in CIProblem is a flaky test in CIoncall: quantizationQuantization support in PyTorchQuantization support in PyTorchtriagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module
Description
Apr 06 22:13:50 test_qadd_relu_same_qparams (__main__.TestQuantizedOps) ... ok (0.008s)
Apr 06 22:13:54 test_qadd_scalar_relu (__main__.TestQuantizedOps) ... Falsifying example: test_qadd_scalar_relu(
Apr 06 22:13:54 self=<__main__.TestQuantizedOps testMethod=test_qadd_scalar_relu>,
Apr 06 22:13:54 A=(array([-418627.78, -500000.38], dtype=float32),
Apr 06 22:13:54 (1960.7857843137256, 255, torch.quint8)),
Apr 06 22:13:54 b=838235.9375,
Apr 06 22:13:54 )
Apr 06 22:13:54 FAIL (4.204s)
Apr 06 22:13:56 test_qclamp (__main__.TestQuantizedOps) ... ok (1.510s)
Apr 06 22:13:57 test_qelu (__main__.TestQuantizedOps) ... ok (0.941s)
Apr 06 22:13:58 test_qhardsigmoid (__main__.TestQuantizedOps) ... ok (0.724s)
Apr 06 22:13:58 test_qmul_broadcast (__main__.TestQuantizedOps) ... ok (0.005s)
Apr 06 22:13:58 test_qmul_relu_different_qparams (__main__.TestQuantizedOps) ... ok (0.008s)
Apr 06 22:14:09 test_qmul_relu_same_qparams (__main__.TestQuantizedOps) ... ok (11.121s)
Apr 06 22:14:09 test_qrelu (__main__.TestQuantizedOps) ... ok (0.545s)
Apr 06 22:14:10 test_qrelu6 (__main__.TestQuantizedOps) ... ok (0.556s)
Apr 06 22:14:11 test_qrelu_leaky (__main__.TestQuantizedOps) ... ok (0.853s)
Apr 06 22:14:12 test_qsigmoid (__main__.TestQuantizedOps) ... ok (1.073s)
Apr 06 22:14:13 test_qtanh (__main__.TestQuantizedOps) ... ok (0.793s)
Apr 06 22:14:13 test_qtopk (__main__.TestQuantizedOps) ... quantization/test_quantized.py:1112: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
Apr 06 22:14:13 indices = torch.tensor(torch.from_numpy(X)).long()
Apr 06 22:14:15 ok (2.559s)
Apr 06 22:14:15 test_qtopk_nhwc (__main__.TestQuantizedOps) ... quantization/test_quantized.py:1138: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
Apr 06 22:14:15 indices = torch.tensor(torch.from_numpy(X)).long()
Apr 06 22:14:19 ok (4.218s)
Apr 06 22:14:19
Apr 06 22:14:19 ======================================================================
Apr 06 22:14:19 FAIL [4.204s]: test_qadd_scalar_relu (__main__.TestQuantizedOps)
Apr 06 22:14:19 ----------------------------------------------------------------------
Apr 06 22:14:19 Traceback (most recent call last):
Apr 06 22:14:19 File "quantization/test_quantized.py", line 399, in test_qadd_scalar_relu
Apr 06 22:14:19 elements=hu.floats(-1e6, 1e6, allow_nan=False),
Apr 06 22:14:19 File "/Users/distiller/workspace/miniconda3/lib/python3.7/site-packages/hypothesis/core.py", line 1090, in wrapped_test
Apr 06 22:14:19 raise the_error_hypothesis_found
Apr 06 22:14:19 File "quantization/test_quantized.py", line 423, in test_qadd_scalar_relu
Apr 06 22:14:19 {} vs {}".format(C_ref.dequantize(), C_hat.dequantize()))
Apr 06 22:14:19 File "/Users/distiller/workspace/miniconda3/lib/python3.7/site-packages/torch/testing/_internal/common_utils.py", line 894, in assertEqual
Apr 06 22:14:19 assertTensorsEqual(x, y)
Apr 06 22:14:19 File "/Users/distiller/workspace/miniconda3/lib/python3.7/site-packages/torch/testing/_internal/common_utils.py", line 860, in assertTensorsEqual
Apr 06 22:14:19 self.assertLessEqual(max_err, prec, message)
Apr 06 22:14:19 AssertionError: tensor(3291.0312) not less than or equal to 1e-05 : Scalar add results don't match:
cc @ezyang @gchanan @zou3519 @jerryzh168 @jianyuh @dzhulgakov @raghuramank100 @jamesr66a
Metadata
Metadata
Assignees
Labels
module: flaky-testsProblem is a flaky test in CIProblem is a flaky test in CIoncall: quantizationQuantization support in PyTorchQuantization support in PyTorchtriagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate module