From 3fbb49e75dc3e7d1d5b14e8a18e983ef9e04ae0b Mon Sep 17 00:00:00 2001
From: Eddie Ren <edwardren@fb.com>
Date: Fri, 10 Sep 2021 12:31:27 -0700
Subject: [PATCH] Extend 2Dim embedding bag benchmarking to include 3Dim
 benchmarks (#64647)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64647

Add support for benchmarking of 8 bit quantizations of N-D batched embeddings. Currently only works for 3Dim embeddings and still requires thought on ramping up from 3Dim to NDim.

Test Plan: ```buck run //caffe2/benchmarks/operator_benchmark/pt:qembedding_pack_test```

Reviewed By: jingsh

Differential Revision: D30770085

fbshipit-source-id: 26659020f3458991592065a05366bde0f060494e
---
 .../operator_benchmark/pt/qembedding_pack_test.py  | 33 ++++++++++++++++++++++
 1 file changed, 33 insertions(+)

diff --git a/benchmarks/operator_benchmark/pt/qembedding_pack_test.py b/benchmarks/operator_benchmark/pt/qembedding_pack_test.py
index f9a3aaf..d55c5fc 100644
--- a/benchmarks/operator_benchmark/pt/qembedding_pack_test.py
+++ b/benchmarks/operator_benchmark/pt/qembedding_pack_test.py
@@ -15,6 +15,13 @@ embeddingbag_conversion_long_configs = op_bench.cross_product_configs(
     tags=('long',)
 )
 
+embeddingbag_conversion_three_dim_configs = op_bench.cross_product_configs(
+    num_embeddings=(80,),
+    embedding_dim=(128, 256, 512),
+    batch_size=(10,),
+    tags=('short',)
+)
+
 conversion_ops = op_bench.op_list(
     attrs=(
         ('qembeddingbag_byte_prepack', torch.ops.quantized.embedding_bag_byte_prepack),
@@ -44,6 +51,16 @@ class EmbeddingBagFloatToFusedBase(op_bench.TorchBenchmarkBase):
     def forward(self, weight):
         return self.op_func(weight)
 
+class EmbeddingBagThreeDimFloatToFusedBase(op_bench.TorchBenchmarkBase):
+    def init(self, num_embeddings, embedding_dim, batch_size, op_func):
+        self.inputs = {
+            "weight": torch.rand(batch_size, num_embeddings, embedding_dim, dtype=torch.float) + 1
+        }
+        self.op_func = op_func
+
+    def forward(self, weight):
+        return self.op_func(weight)
+
 class EmbeddingBagFusedToFloatBase(op_bench.TorchBenchmarkBase):
     def init(self, num_embeddings, embedding_dim, op_func):
         weight = torch.randn(num_embeddings, embedding_dim + 8, dtype=torch.float)
@@ -55,6 +72,16 @@ class EmbeddingBagFusedToFloatBase(op_bench.TorchBenchmarkBase):
     def forward(self, packed_weight):
         return self.op_func(packed_weight)
 
+class EmbeddingBagThreeDimFusedToFloatBase(op_bench.TorchBenchmarkBase):
+    def init(self, num_embeddings, embedding_dim, batch_size, op_func):
+        weight = torch.randn(batch_size, num_embeddings, embedding_dim + 8, dtype=torch.float)
+        self.inputs = {
+            "packed_weight": weight.to(torch.uint8)
+        }
+        self.op_func = op_func
+
+    def forward(self, packed_weight):
+        return self.op_func(packed_weight)
 
 op_bench.generate_pt_tests_from_op_list(conversion_ops,
                                         embeddingbag_conversion_short_configs + embeddingbag_conversion_long_configs,
@@ -62,6 +89,12 @@ op_bench.generate_pt_tests_from_op_list(conversion_ops,
 op_bench.generate_pt_tests_from_op_list(unpack_ops,
                                         embeddingbag_conversion_short_configs + embeddingbag_conversion_long_configs,
                                         EmbeddingBagFusedToFloatBase)
+op_bench.generate_pt_tests_from_op_list(conversion_ops,
+                                        embeddingbag_conversion_three_dim_configs,
+                                        EmbeddingBagThreeDimFloatToFusedBase)
+op_bench.generate_pt_tests_from_op_list(unpack_ops,
+                                        embeddingbag_conversion_three_dim_configs,
+                                        EmbeddingBagThreeDimFusedToFloatBase)
 
 if __name__ == "__main__":
     op_bench.benchmark_runner.main()
-- 
2.7.4