[WebNN EP] Support Einsum op (#19558)

Adds support for einsum via WebNN matmul, transpose, reshape, reducesum, identity and element-wise binary ops.
microsoft · Nov 16, 2024 · 5928009 · 5928009
1 parent c73a3d1
commit 5928009
Show file tree

Hide file tree

Showing 5 changed files with 800 additions and 0 deletions.
diff --git a/js/web/docs/webnn-operators.md b/js/web/docs/webnn-operators.md
@@ -29,6 +29,7 @@ operators and the supported opset domain/versions in **WebNN EP** by ONNX Runtim
 | Div | ai.onnx(7-12, 13, 14+) | div | ✓ | ✓ | |
 | DequantizeLinear | ai.onnx(10-12, 13-18, 19-20, 21-22, 23+) | dequantizeLinear | ✗ | ✓ | |
 | Dropout | ai.onnx(7-9, 10-11, 12, 13-21, 22+) | identity | ✓ | ✓ | Only supports test mode |
+| Einsum | ai.onnx(12+) | reshape, transpose, matmul, reduceSum, mul, triangular | ✓ | ✓ | |
 | Elu | ai.onnx(7+) | elu | ✓ | ✓ | WebNN CPU backend only supports 'alpha' value is 1.0 |
 | Equal | ai.onnx(7-10, 11-12, 13-18, 19+) | equal | ✓ | ✓ | |
 | Erf | ai.onnx(7-9, 10-12, 13+) | erf | ✓ | ✓ | |

diff --git a/onnxruntime/core/providers/webnn/builders/helper.h b/onnxruntime/core/providers/webnn/builders/helper.h
@@ -210,6 +210,7 @@ static const InlinedHashMap<std::string, std::string> op_map = {
     {"DequantizeLinear", "dequantizeLinear"},
     {"Dropout", "identity"},
     {"DynamicQuantizeLinear", "dynamicQuantizeLinear"},
+    {"Einsum", "matmul"},
     {"Elu", "elu"},
     {"Equal", "equal"},
     {"Erf", "erf"},