[SOT] Ensure run AST transformed function under graph_tracing_guard

[DrRyanHuang]

PR Category

Execute Infrastructure

PR Types

Bug fixes

Description

动转静会在进入 run_program 之前，将所有的 Tensor，统一转化为 pir.Value

import paddle
from paddle import nn

class Net(nn.Layer):
    def __init__(self):
        super().__init__()
        self.layer = nn.Linear(8, 8)

    @paddle.jit.marker.capture_control_flow
    def fn(self, x):
        if x.sum() > 0:
            x = self.layer(x)
            x += paddle.ones_like(x)
        else:
            x = self.layer(x)
            x -= paddle.zeros_like(x)
        return x

    @paddle.jit.to_static(backend=None)
    def forward(self, x):
        return self.fn(x) + 1

if __name__ == "__main__":
    paddle.seed(101)
    model = Net()
    print("model", model)
    input_data = paddle.rand([2, 8])
    output = model(input_data)
    print(output)

上述例子中的，model.layer.weight 存在未被转化的情况

TODO:

• 后续请 @SigureMo 分享一下动转静 Tensor 转化为 pir.Value 的策略喵~

[paddle-bot]

你的PR提交成功，感谢你对开源项目的贡献!
请关注后续CI自动化测试结果，详情请参考Paddle-CI手册。
Your PR has been submitted. Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
⚠️ Please upload report for BASE (develop@8e2251f). Learn more about missing BASE report.

Additional details and impacted files

@@             Coverage Diff             @@
##             develop    #76198   +/-   ##
===========================================
  Coverage           ?   100.00%           
===========================================
  Files              ?         5           
  Lines              ?        31           
  Branches           ?         0           
===========================================
  Hits               ?        31           
  Misses             ?         0           
  Partials           ?         0           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.