chore: import upstream snapshot with attribution

test/tensorrt/get_program.py

@@ -0,0 +1,226 @@
+# Copyright (c) 2024 PaddlePaddle Authors. All Rights Reserved
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+import numpy as np
+
+import paddle
+from paddle import nn, static
+from paddle.nn import TransformerEncoder, TransformerEncoderLayer
+
+
+def get_r50_program():
+    paddle.enable_static()
+    from paddle.vision.models import wide_resnet50_2
+
+    with paddle.pir_utils.IrGuard():
+        infer_program = paddle.static.Program()
+        startup_program = paddle.static.Program()
+        with static.program_guard(infer_program, startup_program):
+            scope = paddle.static.global_scope()
+            input_data = paddle.static.data(
+                shape=[-1, 3, 224, 224], dtype='float32', name='input'
+            )
+            model = wide_resnet50_2()
+            model.eval()
+            output = model(input_data)
+        place = paddle.CUDAPlace(0)
+        exe = static.Executor(place)
+        exe.run(startup_program)
+
+    params = infer_program.global_block().all_parameters()
+    param_dict = {}
+    for v in params:
+        name = v.get_defining_op().attrs()["parameter_name"]
+        param_dict.update({name: np.array(scope.var(name).get_tensor())})
+
+    return infer_program, scope, param_dict
+
+
+def get_r50_refit_program(save_path):
+    paddle.enable_static()
+    from paddle.vision.models import wide_resnet50_2
+
+    infer_program = paddle.static.Program()
+    startup_program = paddle.static.Program()
+    with paddle.static.program_guard(infer_program, startup_program):
+        scope = paddle.static.global_scope()
+        input_data = paddle.static.data(
+            shape=[-1, 3, 224, 224], dtype='float32', name='input'
+        )
+        model = wide_resnet50_2()
+        model.eval()
+        output = model(input_data)
+
+        place = paddle.CUDAPlace(0)
+        exe = paddle.static.Executor(place)
+        exe.run(startup_program)
+        _ = exe.run(
+            infer_program,
+            feed={'input': np.random.randn(1, 3, 224, 224).astype(np.float32)},
+            fetch_list=[output],
+        )
+
+    paddle.static.save_inference_model(
+        path_prefix=save_path,
+        feed_vars=[input_data],
+        fetch_vars=[output],
+        executor=exe,
+        program=infer_program,
+    )
+
+    params = infer_program.global_block().all_parameters()
+    param_dict = {}
+    for v in params:
+        name = v.get_defining_op().attrs()["parameter_name"]
+        param_dict.update({name: np.array(scope.var(name).get_tensor())})
+
+    return infer_program, scope, param_dict
+
+
+def get_dummy_program():
+    paddle.enable_static()
+    with paddle.pir_utils.IrGuard():
+        main_program = paddle.static.Program()
+        default_startup_program = paddle.static.Program()
+        with paddle.static.program_guard(main_program, default_startup_program):
+            scope = paddle.static.global_scope()
+            input = paddle.static.data(
+                shape=[-1, 64], dtype='float32', name='input'
+            )
+            weight_numpy = np.random.rand(64, 64).astype('float32')
+            weight = paddle.create_parameter(
+                name="w",
+                shape=[64, 64],
+                dtype='float32',
+                attr=paddle.ParamAttr(
+                    initializer=paddle.nn.initializer.Assign(weight_numpy)
+                ),
+            )
+            bias_numpy = np.random.rand(64).astype('float32')
+            bias = paddle.create_parameter(
+                name="b",
+                shape=[64],
+                dtype='float32',
+                attr=paddle.ParamAttr(
+                    initializer=paddle.nn.initializer.Assign(bias_numpy)
+                ),
+            )
+            x = paddle.matmul(input, weight)
+            x_1 = paddle.add(x, bias)
+            x_1 = paddle.unsqueeze(x_1, axis=0)
+            x_1 = paddle.squeeze(x_1, axis=0)
+            y = paddle.nn.functional.relu(x_1)
+            y_gelu_1 = paddle.nn.functional.gelu(y)
+            y_gelu_2 = paddle.nn.functional.gelu(x_1)
+
+            # Concatenate the outputs of the two GELU operations
+            concat_out = paddle.concat([y_gelu_1, y_gelu_2], axis=-1)
+            output = paddle.unsqueeze(concat_out, axis=0)
+
+        exe = paddle.static.Executor(paddle.CUDAPlace(0))
+        exe.run(default_startup_program)
+
+        params = main_program.global_block().all_parameters()
+        param_dict = {}
+        # save parameters
+        for v in params:
+            name = v.get_defining_op().attrs()["parameter_name"]
+            param_dict.update({name: np.array(scope.var(name).get_tensor())})
+    return main_program, scope, param_dict
+
+
+class BertModel(nn.Layer):
+    def __init__(
+        self,
+        vocab_size,
+        hidden_size=768,
+        num_hidden_layers=12,
+        num_attention_heads=12,
+    ):
+        super().__init__()
+        self.embeddings = nn.Embedding(vocab_size, hidden_size)
+        encoder_layer = TransformerEncoderLayer(
+            hidden_size, num_attention_heads, hidden_size * 4
+        )
+        self.encoder = TransformerEncoder(encoder_layer, num_hidden_layers)
+
+    def forward(self, input_ids):
+        embeddings = self.embeddings(input_ids)
+        encoded_output = self.encoder(embeddings)
+        return encoded_output
+
+
+def get_bert_program():
+    paddle.enable_static()
+
+    vocab_size = 30522  # BERT base vocab size
+    hidden_size = 768
+    num_hidden_layers = 2
+    num_attention_heads = 12
+    seq_length = 128
+
+    with paddle.pir_utils.IrGuard():
+        main_program = static.default_main_program()
+        startup_program = static.default_startup_program()
+        with static.program_guard(main_program, startup_program):
+            scope = paddle.static.global_scope()
+            input_ids = static.data(
+                name='input_ids', shape=[-1, -1], dtype='int64'
+            )
+
+            bert_model = BertModel(
+                vocab_size, hidden_size, num_hidden_layers, num_attention_heads
+            )
+            bert_model.eval()
+            logits = bert_model(input_ids)
+
+    place = (
+        paddle.CUDAPlace(0)
+        if paddle.is_compiled_with_cuda()
+        else paddle.CPUPlace()
+    )
+    pir_program = main_program
+
+    with (
+        paddle.pir_utils.IrGuard(),
+        paddle.static.program_guard(pir_program, startup_program),
+    ):
+        x = np.ones([1, seq_length]).astype('int64')
+        executor = paddle.static.Executor(place)
+        executor.run(startup_program)
+        fetches = executor.run(
+            pir_program,
+            feed={"input_ids": x},
+            fetch_list=pir_program.list_vars()[-3],
+        )
+
+    params = main_program.global_block().all_parameters()
+    param_dict = {}
+    # save parameters
+    for v in params:
+        name = v.get_defining_op().attrs()["parameter_name"]
+        param_dict.update({name: np.array(scope.var(name).get_tensor())})
+    return pir_program, scope, param_dict
+
+
+class SimpleGatherNet(nn.Layer):
+    def __init__(self):
+        super().__init__()
+        self.linear = paddle.nn.Linear(149600, 1)
+
+    def forward(self, map_vector_features, polyline_mask):
+        map_vector_features = map_vector_features[polyline_mask]
+
+        return map_vector_features