[NewIR]Remove compatible logic of ProgramTranslator (#55453)

paddle/fluid/ir_adaptor/translator/op_compat_gen.py

@@ -126,9 +126,6 @@ def insert_new_mutable_attributes(
                     backward_op, op_compat_item["scalar"]
                 )
 
-    # special op mappings
-    op_name_mappings["fetch_v2"] = "fetch"
-
     op_name_normailzer_template = env.get_template("op_compat_info.cc.j2")
     with open(output_source_file, 'wt') as f:
         op_compat_definition = op_name_normailzer_template.render(

paddle/fluid/ir_adaptor/translator/op_compat_info.h

@@ -19,8 +19,6 @@
 
 #include "glog/logging.h"
 
-#include "paddle/fluid/ir_adaptor/translator/utils.h"
-
 #pragma once
 
 namespace paddle {
@@ -106,11 +104,11 @@ class OpNameNormalizer {
       return legacy_name;
     }
     if (op_arg_name_mappings.find(op_type) == op_arg_name_mappings.end()) {
-      return UnderscoreToCamelCase(arg_name);
+      return arg_name;
     }
     auto& arg_mappings = op_arg_name_mappings[op_type];
     if (arg_mappings.find(arg_name) == arg_mappings.end()) {
-      return UnderscoreToCamelCase(arg_name);
+      return arg_name;
     }
     return arg_mappings.at(arg_name);
   }

paddle/fluid/ir_adaptor/translator/op_translator.cc

@@ -307,7 +307,7 @@ ir::OpInfo OpTranscriber::LoopkUpOpInfo(ir::IrContext* ctx,
                                         const OpDesc& op_desc) {
   std::string target_op_name =
       kTargetDialectPrefix + OpNameCompatibleMapping(op_desc.Type());
-  if (IsInplace(op_desc)) {
+  if (IsInplace(op_desc) && *target_op_name.rbegin() != '_') {
     target_op_name += "_";
   }
   VLOG(6) << "[op name normalizing]: " << op_desc.Type() << " to "

paddle/phi/api/yaml/op_compat.yaml

@@ -354,6 +354,7 @@
     attrs : [bool use_mkldnn = false]
 
 - op : bilinear (bilinear_tensor_product)
+  backward: bilinear_grad (bilinear_tensor_product_grad)
   inputs :
     {x : X, y : Y,weight: Weight, bias: Bias}
   outputs :
@@ -1838,7 +1839,7 @@
       data_type : float
       support_tensor : true
 
-- op : merged_momentum_
+- op : merged_momentum_ (merged_momentum)
   inputs :
     {param : Param, grad : Grad, velocity : Velocity, learning_rate : LearningRate, master_param : MasterParam}
   outputs :
@@ -3038,11 +3039,27 @@
     yolo_loss : GetYoloLossExpectedKernelType
     yolo_loss_grad : GetYoloLossExpectedKernelType
 
-- op: fetch
+- op: channel_shuffle
+  inputs:
+    {x: X}
+  outputs:
+    {out: Out}
+
+- op: fetch (fetch_v2)
   inputs: {x: X}
   outputs: {out: Out}
 
 - op: full_batch_size_like (fill_constant_batch_size_like)
+  inputs:
+    {input: Input}
+  outputs:
+    {out: Out}
+
+- op: logspace
+  inputs:
+    {start: Start, stop: Stop, num: Num, base: Base}
+  outputs:
+    {out: Out}
 
 - op: lu
   backward: lu_grad
@@ -3059,6 +3076,12 @@
   outputs :
     {reindex_src : Reindex_Src, reindex_dst : Reindex_Dst, out_nodes : Out_Nodes}
 
+- op: rrelu
+  inputs:
+    {x: X}
+  outputs:
+    {out: Out, noise: Noise}
+
 - op: sigmoid_cross_entropy_with_logits
   backward: sigmoid_cross_entropy_with_logits_grad
   inputs :

test/ir/new_ir/test_special_op_translator.py

@@ -194,5 +194,21 @@ def test_with_axis(self):
                 np.testing.assert_array_equal(out[0], np.all(arr, axis=0))
 
 
+class TestIndexPutOpTranscriber(unittest.TestCase):
+    def test_op(self):
+        place = core.Place()
+        place.set_place(paddle.CPUPlace())
+        new_scope = paddle.static.Scope()
+        main_program = paddle.static.Program()
+        with paddle.static.scope_guard(new_scope):
+            with paddle.static.program_guard(main_program):
+                x = paddle.randn([2, 3])
+                indices = [paddle.randint(0, 2, [2]), paddle.randint(0, 1, [2])]
+                value = paddle.randn([2])
+                y = paddle.index_put(x, indices, value, False)
+
+        _ = ir.translate_to_new_ir(main_program.desc)
+
+
 if __name__ == "__main__":
     unittest.main()