[PIR] Add Op output type check for CSE

[SigureMo]

PR Category

Execute Infrastructure

PR Types

Bug fixes

Description

CSE 的 equal 判断里添加了对 OP 输出 type 的检查

如下所示，这些 assign_value attributes 都是一样的，但是输出 type 不一样，按理说正常组网 infermeta 是不会出现这种情况，这个模型是 PT 转换得来，可能是 PT 或者老 IR 组网的时候哪里有 trick 对输出 type 进行了重新 set

    (%1892) = "pd_op.assign_value" () {dtype:(pd_op.DataType)float32,persistable:[false],place:(pd_op.Place)Place(undefined:0),shape:[(Int32)1],stop_gradient:[false],values:[(Double)1.77113e+27]} : () -> builtin.tensor<1xf32>
    (%1893) = "pd_op.assign_value" () {dtype:(pd_op.DataType)float32,persistable:[false],place:(pd_op.Place)Place(undefined:0),shape:[(Int32)1],stop_gradient:[false],values:[(Double)1.77113e+27]} : () -> builtin.tensor<2000xi64>
    (%1894) = "pd_op.assign_value" () {dtype:(pd_op.DataType)float32,persistable:[false],place:(pd_op.Place)Place(undefined:0),shape:[(Int32)1],stop_gradient:[false],values:[(Double)1.77113e+27]} : () -> builtin.tensor<1xf32>
    (%1895) = "pd_op.assign_value" () {dtype:(pd_op.DataType)float32,persistable:[false],place:(pd_op.Place)Place(undefined:0),shape:[(Int32)1],stop_gradient:[false],values:[(Double)1.77113e+27]} : () -> builtin.tensor<-1xi32>
    (%1896) = "pd_op.assign_value" () {dtype:(pd_op.DataType)float32,persistable:[false],place:(pd_op.Place)Place(undefined:0),shape:[(Int32)1],stop_gradient:[false],values:[(Double)1.77113e+27]} : () -> builtin.tensor<-1x10xf32>
    (%1897) = "pd_op.assign_value" () {dtype:(pd_op.DataType)float32,persistable:[false],place:(pd_op.Place)Place(undefined:0),shape:[(Int32)1],stop_gradient:[false],values:[(Double)1.77113e+27]} : () -> builtin.tensor<1xi64>
    (%1898) = "pd_op.assign_value" () {dtype:(pd_op.DataType)float32,persistable:[false],place:(pd_op.Place)Place(undefined:0),shape:[(Int32)1],stop_gradient:[false],values:[(Double)1.77113e+27]} : () -> builtin.tensor<-1x5xf32>
    (%1899) = "pd_op.assign_value" () {dtype:(pd_op.DataType)float32,persistable:[false],place:(pd_op.Place)Place(undefined:0),shape:[(Int32)1],stop_gradient:[false],values:[(Double)1.77113e+27]} : () -> builtin.tensor<-1x5xf32>
    (%1900) = "pd_op.assign_value" () {dtype:(pd_op.DataType)float32,persistable:[false],place:(pd_op.Place)Place(undefined:0),shape:[(Int32)1],stop_gradient:[false],values:[(Double)1.77113e+27]} : () -> builtin.tensor<-1x5xf32>
    (%1901) = "pd_op.assign_value" () {dtype:(pd_op.DataType)float32,persistable:[false],place:(pd_op.Place)Place(undefined:0),shape:[(Int32)1],stop_gradient:[false],values:[(Double)1.77113e+27]} : () -> builtin.tensor<-1x5xf32>
    (%1902) = "pd_op.assign_value" () {dtype:(pd_op.DataType)float32,persistable:[false],place:(pd_op.Place)Place(undefined:0),shape:[(Int32)1],stop_gradient:[false],values:[(Double)1.77113e+27]} : () -> builtin.tensor<-1x5xf32>
    (%1903) = "pd_op.assign_value" () {dtype:(pd_op.DataType)float32,persistable:[false],place:(pd_op.Place)Place(undefined:0),shape:[(Int32)1],stop_gradient:[false],values:[(Double)1.77113e+27]} : () -> builtin.tensor<-1x9xf32>
    (%1904) = "pd_op.assign_value" () {dtype:(pd_op.DataType)float32,persistable:[false],place:(pd_op.Place)Place(undefined:0),shape:[(Int32)1],stop_gradient:[false],values:[(Double)1.77113e+27]} : () -> builtin.tensor<-1x9xf32>
    (%1905) = "pd_op.assign_value" () {dtype:(pd_op.DataType)float32,persistable:[false],place:(pd_op.Place)Place(undefined:0),shape:[(Int32)1],stop_gradient:[false],values:[(Double)1.77113e+27]} : () -> builtin.tensor<-1x9xf32>
    (%1906) = "pd_op.assign_value" () {dtype:(pd_op.DataType)float32,persistable:[false],place:(pd_op.Place)Place(undefined:0),shape:[(Int32)1],stop_gradient:[false],values:[(Double)1.77113e+27]} : () -> builtin.tensor<-1x9xf32>
    (%1907) = "pd_op.assign_value" () {dtype:(pd_op.DataType)float32,persistable:[false],place:(pd_op.Place)Place(undefined:0),shape:[(Int32)1],stop_gradient:[false],values:[(Double)1.77113e+27]} : () -> builtin.tensor<-1x9xf32>

PCard-66972

[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.