Add Split QuickGelu Fusion #20344
Add Split QuickGelu Fusion #20344
Conversation
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
| // Trying to find Split->QuickGelu->Mul Path | ||
| std::vector<const Node::EdgeEnd*> edges; | ||
| std::vector<graph_utils::EdgeEndToMatch> quickgelu_mul_path{ | ||
| {0, 0, "QuickGelu", {1, 11, 13}, kOnnxDomain}, |
There was a problem hiding this comment.
the activation can be other elementwise operator instead of QuickGelu, maybe we can add a todo here to support other operators.
| "fused " + split_node.Name() + " and " + quickgelu_node.Name() + " and " + mul.Name() + " into SplitQuickGelu"; | ||
|
|
||
| std::string op_type = "SplitQuickGelu"; | ||
| Node& fused_node = graph.AddNode(graph.GenerateNodeName(op_type), |
There was a problem hiding this comment.
nit: add transformer name into node's name to make future debugging easier to figure out which graph transformer generate the node.
…/add_split_quickgelu_fusion
| @@ -0,0 +1,38 @@ | |||
| import onnx | |||
Check notice
Code scanning / CodeQL
Module is imported with 'import' and 'import from' Note test
| @@ -0,0 +1,50 @@ | |||
| import torch | |||
Check warning
Code scanning / lintrunner
BLACK-ISORT/format Warning test
| # inp = torch.randn((76, 54, 1368), device='cuda:0') | ||
| # out = temp(inp) | ||
|
|
||
| from onnxruntime.training.ortmodule import ORTModule |
Check warning
Code scanning / lintrunner
RUFF/E402 Warning test
| # out = temp(inp) | ||
|
|
||
| from onnxruntime.training.ortmodule import ORTModule | ||
| import time |
Check warning
Code scanning / lintrunner
RUFF/E402 Warning test
|
|
||
| def compare_torch_ort_perf(inp, num_steps=1000): | ||
| model2 = ORTModule(temp) | ||
| ort_temp_out = model2(inp) |
Check warning
Code scanning / lintrunner
RUFF/F841 Warning test
| print("Current input shape:", inp.shape) | ||
| torch.cuda.synchronize() | ||
| start = time.time() | ||
| for i in range(num_steps): |
Check warning
Code scanning / lintrunner
RUFF/B007 Warning test
| print("Total time torch:", time.time() - start) | ||
| torch.cuda.synchronize() | ||
| start = time.time() | ||
| for i in range(num_steps): |
Check warning
Code scanning / lintrunner
RUFF/B007 Warning test
|
|
||
| def compare_torch_ort_perf(inp, num_steps=1000): | ||
| model2 = ORTModule(temp) | ||
| ort_temp_out = model2(inp) |
Check notice
Code scanning / CodeQL
Unused local variable Note test
Description
Add Split QuickGelu Fusion
Motivation and Context