From 334339a3d27d17d955c63b97f32dd728d7212f8c Mon Sep 17 00:00:00 2001
From: Philip Meier <github.pmeier@posteo.de>
Date: Tue, 8 Feb 2022 14:36:26 +0100
Subject: [PATCH] add `OpInfo`s for
 `torch.nn.functional.triplet_margin(_with_distance)?_loss`

ghstack-source-id: bbc38b4b859dd116bf604ae5a44ed40d642005b1
Pull Request resolved: https://github.com/pytorch/pytorch/pull/67079
---
 .../_internal/common_methods_invocations.py   | 50 +++++++++++++++++++
 1 file changed, 50 insertions(+)

diff --git a/torch/testing/_internal/common_methods_invocations.py b/torch/testing/_internal/common_methods_invocations.py
index cc089a3a5d0..0c8b6f85c1b 100644
--- a/torch/testing/_internal/common_methods_invocations.py
+++ b/torch/testing/_internal/common_methods_invocations.py
@@ -10,6 +10,7 @@ import unittest
 import math
 
 import torch
+from torch import nn
 import numpy as np
 from torch._six import inf
 import collections.abc
@@ -7770,6 +7771,25 @@ def sample_inputs_binary_cross_entropy(op_info, device, dtype, requires_grad, lo
         for shape, kwargs in shapes_and_kwargs
     ]
 
+def sample_inputs_triplet_margin_loss(op_info, device, dtype, requires_grad, with_distance=False, **kwargs):
+    make = partial(make_tensor, (S, M), device=device, dtype=dtype, requires_grad=requires_grad)
+
+    kwargss = (
+        *[dict(margin=margin) for margin in (1e-6, 1.0, 10.0)],
+        dict(swap=True),
+        *[dict(reduction=reduction) for reduction in ("mean", "sum", "none")],
+    )
+
+    sample_inputs = []
+    for kwargs in kwargss:
+        input = make()
+        args = (make(), make())
+        if with_distance:
+            kwargs["distance_function"] = nn.PairwiseDistance()
+        sample_inputs.append(SampleInput(input, args=args, kwargs=kwargs))
+
+    return sample_inputs
+
 
 foreach_unary_op_db: List[OpInfo] = [
     ForeachFuncInfo('exp'),
@@ -15702,6 +15722,36 @@ op_db: List[OpInfo] = [
             ),
         ),
     ),
+    OpInfo(
+        "nn.functional.triplet_margin_loss",
+        sample_inputs_func=sample_inputs_triplet_margin_loss,
+        dtypes=all_types_and_complex_and(torch.bfloat16),
+        dtypesIfCUDA=all_types_and_complex_and(torch.float16, torch.bfloat16),
+        supports_out=False,
+    ),
+    OpInfo(
+        "nn.functional.triplet_margin_with_distance_loss",
+        sample_inputs_func=partial(sample_inputs_triplet_margin_loss, with_distance=True),
+        dtypes=all_types_and_complex_and(torch.bfloat16),
+        dtypesIfCUDA=all_types_and_complex_and(torch.float16, torch.bfloat16),
+        supports_out=False,
+        skips=(
+            # This test cannot handle a callable passed to `distance_function`. If we would use
+            # `distance_function=None`, the test would pass fine.
+            DecorateInfo(
+                unittest.expectedFailure,
+                "TestJit",
+                "test_variant_consistency_jit",
+            ),
+            # This tests raises a plain AssertionError at
+            # https://github.com/pytorch/pytorch/blob/840fe8e4e6efee5d8197bd6987757f29d72dd162/test/test_fx_experimental.py#L1559
+            DecorateInfo(
+                unittest.expectedFailure,
+                "TestNormalizeOperators",
+                "test_normalize_operator_exhaustive",
+            ),
+        ),
+    ),
 ]
 
 # Common operator groupings