[Paddle Inference] Add add eye trt converter #48937
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
你的PR提交成功，感谢你对开源项目的贡献! |
zhangjun
reviewed
Dec 9, 2022
| // Declare inputs attr | ||
| const int num_rows = PADDLE_GET_CONST(int, op_desc.GetAttr("num_rows")); | ||
| int num_columns = PADDLE_GET_CONST(int, op_desc.GetAttr("num_columns")); | ||
| const int dtype = PADDLE_GET_CONST(int, op_desc.GetAttr("dtype")); |
There was a problem hiding this comment.
可以改成
auto dtype = static_cast<framework::proto::VarType::Type>(
PADDLE_GET_CONST(int, op_desc.GetAttr("dtype")));
Comment on lines
79
to
82
| if (-1 == num_columns) { | ||
| input_shape.d[1] = num_rows; | ||
| num_columns = num_rows; | ||
| } |
There was a problem hiding this comment.
num_columns为默认值-1时，num_columns=num_rows, 这段逻辑可以放在input_shape.d赋值前
| } | ||
|
|
||
| std::vector<T> constant_arr(num_rows * num_columns, 0); | ||
| for (int i = 0; i < std::min(num_rows, num_columns); i++) { |
There was a problem hiding this comment.
std::min(num_rows, num_columns) 放循环外避免多次调用std::min
Comment on lines
52
to
72
| nvinfer1::DataType nv_type = nvinfer1::DataType::kFLOAT; | ||
| switch (dtype) { | ||
| case paddle::framework::proto::VarType::FP32: | ||
| nv_type = nvinfer1::DataType::kFLOAT; | ||
| typedef float T; | ||
| break; | ||
| case paddle::framework::proto::VarType::FP16: | ||
| nv_type = nvinfer1::DataType::kHALF; | ||
| typedef uint16_t T; | ||
| break; | ||
| case paddle::framework::proto::VarType::INT32: | ||
| nv_type = nvinfer1::DataType::kINT32; | ||
| typedef int32_t T; | ||
| break; | ||
| default: | ||
| paddle::platform::errors::InvalidArgument( | ||
| "Paddle-TRT loads weighths failed, found not supported data type " | ||
| "%s.", | ||
| dtype); | ||
| break; | ||
| } |
There was a problem hiding this comment.
这里T类型声明编译报错，可以参考convert/fill_constant_op.cc 中写法，进行不同类型数组赋值
zhangjun
reviewed
Dec 9, 2022
| for _ in range(6): | ||
| if np.random.random() > 0.5: | ||
| num_rows = generate_input_attr1() | ||
| attr_dic = {"num_rows": num_rows} |
There was a problem hiding this comment.
可以加一个num_columns为-1。两处attr_dic中添加一个dtype字段，37-38行之间添加一层for循环用于dtype赋值，
for dtype in [2, 4, 5]
@zhangjun 打扰了，问一下这个 |
|
很抱歉，经过我们的反复讨论，你的PR暂未达到合入标准，请阅读飞桨原生算子开发规范，你可以重新提交新的PR，我们先将此PR关闭，感谢你的贡献。 |
# for free
to join this conversation on GitHub.
Already have an account?
# to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
PR types
Others
PR changes
Others
Describe
#48292