update path

Author: tp-nan

docs/contribution_guide/modify_the_code.md

@@ -28,5 +28,5 @@ pytest .
 If necessary, please consider supplementing with [Python tests](https://github.com/torchpipe/torchpipe//test).
 
 :::note Code Formatting (optional)
-Please configure a formatting plugin to enable [.clang-format](https://github.com/torchpipe/torchpipe/-/blob/develop/.clang-format).
+Please configure a formatting plugin to enable [.clang-format](https://github.com/torchpipe/torchpipe/blob/develop/.clang-format).
 :::

docs/quick_start_new_user.md

@@ -7,7 +7,7 @@ type: explainer
 
 # Trial in 30mins(new users)
 
-TorchPipe is a multi-instance pipeline parallel library that provides a seamless integration between lower-level acceleration libraries (such as TensorRT and OpenCV) and RPC frameworks. It guarantees high service throughput while meeting latency requirements. This document is mainly for new users, that is, users who are in the introductory stage of acceleration-related theoretical knowledge, know some python grammar, and can read simple codes. This content mainly includes the use of torchpipe for accelerating service deployment, complemented by performance and effect comparisons. The complete code of this document can be found at [resnet50_thrift](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/resnet50_thrift/)。
+TorchPipe is a multi-instance pipeline parallel library that provides a seamless integration between lower-level acceleration libraries (such as TensorRT and OpenCV) and RPC frameworks. It guarantees high service throughput while meeting latency requirements. This document is mainly for new users, that is, users who are in the introductory stage of acceleration-related theoretical knowledge, know some python grammar, and can read simple codes. This content mainly includes the use of torchpipe for accelerating service deployment, complemented by performance and effect comparisons. The complete code of this document can be found at [resnet50_thrift](https://github.com/torchpipe/torchpipe/blob/develop/examples/resnet50_thrift/)。
 
 ## Catalogue
 * [1. Basic knowledge](#1)
@@ -84,7 +84,7 @@ self.classification_engine = torch2trt(resnet50, [input_shape],
 
 ```
 
-The overall online service deployment can be found at [main_trt.py](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/resnet50_thrift/main_trt.py)
+The overall online service deployment can be found at [main_trt.py](https://github.com/torchpipe/torchpipe/blob/develop/examples/resnet50_thrift/main_trt.py)
 
 :::tip
 Since TensorRT is not thread-safe, when using this method for model acceleration, it is necessary to handle locking (with self.lock:) during the service deployment process.
@@ -104,7 +104,7 @@ From the above process, it's clear that when accelerating a single model, the fo
 
 ![](images/quick_start_new_user/torchpipe_en.png)
 
-We've made adjustments to the deployment of our service using TorchPipe.The overall online service deployment can be found at [main_torchpipe.py](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/resnet50_thrift/main_torchpipe.py).
+We've made adjustments to the deployment of our service using TorchPipe.The overall online service deployment can be found at [main_torchpipe.py](https://github.com/torchpipe/torchpipe/blob/develop/examples/resnet50_thrift/main_torchpipe.py).
 The core function modifications as follows:
 
 ```py
@@ -219,7 +219,7 @@ std="58.395, 57.120, 57.375" # 255*"0.229, 0.224, 0.225"
 `python clien_qps.py --img_dir /your/testimg/path/ --port 8888 --request_client 20 --request_batch 1
 `
 
-The specific test code can be found at [client_qps.py](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/resnet50_thrift/client_qps.py)
+The specific test code can be found at [client_qps.py](https://github.com/torchpipe/torchpipe/blob/develop/examples/resnet50_thrift/client_qps.py)
 
 With the same Thrift service interface, testing on a machine with NIDIA-3080 GPU, 36-core CPU, and concurrency of 10, we have the following results:
 

docs/tools/quantization.mdx

@@ -29,7 +29,7 @@ For detection models, you can consider using the [official complete tutorial](ht
 
 In addition to the pre-training parameters provided by the model for normal training, training-based quantization also requires quantization pre-training parameters provided by post-training quantization (ptq).
 
-We have integrated [calib_tools](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/calib_tools.py) for reference.
+We have integrated [calib_tools](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/calib_tools.py) for reference.
 
 - Define calibrator:
 ```python
@@ -100,17 +100,17 @@ The official training format is very simple and is only used as an example.
 #### Direct Quantization without Modifying Backbone
 Following the official example, we conducted step-by-step experiments on resnet:
 
-- Download training data: [code](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/download_data.py)
-- Train for 10 epochs to obtain the resnet50 model: [code](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/fp32_train.py), accuracy 98.44%
-- (optional) PyTorch ptq: [code](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/ptq.py), accuracy 96.64% (max)
-- (optional) PyTorch qat: [code](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/qat.py), accuracy 98.26%.
+- Download training data: [code](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/download_data.py)
+- Train for 10 epochs to obtain the resnet50 model: [code](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/fp32_train.py), accuracy 98.44%
+- (optional) PyTorch ptq: [code](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/ptq.py), accuracy 96.64% (max)
+- (optional) PyTorch qat: [code](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/qat.py), accuracy 98.26%.
 
 #### MSE + Residual Fusion {#mseadd}
 
 The above resnet training uses the max quantization method and does not fuse the Add layer, resulting in TensorRT running speed not meeting expectations. The following are the results after fusing Add under int8 and switching to the mse mode:
 
-- ptq: [code](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/ptq_merge_residual.py), accuracy 94.34% (mse)
-- qat: [code](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/qat_merge_residual.py), accuracy 95.82%.
+- ptq: [code](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/ptq_merge_residual.py), accuracy 94.34% (mse)
+- qat: [code](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/qat_merge_residual.py), accuracy 95.82%.
 
 
 #### Summary of Results in PyTorch
@@ -124,9 +124,9 @@ The above resnet training uses the max quantization method and does not fuse the
 ### Summary of Test Results in TorchPipe
 The following tests were performed using the onnx generated by TorchPipe:
 
-- Export onnx: [code](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/export_onnx_merge_residual.py)
-- Load fp32-onnx with TorchPipe and perform ptq: [code](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/torchpipe_ptq_test.py)
-- Test with qat-onnx loaded with TorchPipe: [code](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/torchpipe_qat_test.py)
+- Export onnx: [code](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/export_onnx_merge_residual.py)
+- Load fp32-onnx with TorchPipe and perform ptq: [code](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/torchpipe_ptq_test.py)
+- Test with qat-onnx loaded with TorchPipe: [code](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/torchpipe_qat_test.py)
 
 
 | Model                  | Accuracy | Performance | Note                                |
@@ -135,4 +135,4 @@ The following tests were performed using the onnx generated by TorchPipe:
 | tensorrt's native int8 | 98.26%   | -           |                                     |
 | qat                    | 98.67%   | -           | [Acc. under onnxruntime] is 98.69%. |
 
-[Acc. under onnxruntime]: https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/onnxruntime_qat_test.py
+[Acc. under onnxruntime]: https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/onnxruntime_qat_test.py

i18n/zh/docusaurus-plugin-content-docs/current/contribution_guide/modify_the_code.md

@@ -28,5 +28,5 @@ pytest .
 需要时请考虑补充[python测试](https://github.com/torchpipe/torchpipe//test)。
 
 :::note 代码格式（optinal）
-请配置格式化插件以便[.clang-format](https://github.com/torchpipe/torchpipe/-/blob/develop/.clang-format)生效。
+请配置格式化插件以便[.clang-format](https://github.com/torchpipe/torchpipe/blob/develop/.clang-format)生效。
 :::

i18n/zh/docusaurus-plugin-content-docs/current/quick_start_new_user.md

@@ -7,7 +7,7 @@ type: explainer
 
 # torchpipe快速上手（30min体验版）
 
-torchpipe是为工业界所准备的一个独立作用于底层加速库（如tensorrt，opencv，torchscript）以及 RPC（如thrift, gRPC）之间的多实例流水线并行库，助力使用者能在部署阶段节约更多的硬件资源，帮助产品应用落地。此教程主要针对初级用户，即对于加速相关的理论知识处于入门阶段，具有一定的 Python基础，能够阅读简单代码的用户。此内容主要包括使用torchpipe进行服务部署加速的使用方法、性能和效果差异对比等。本文档的完整代码见可详见[resnet50_thrift](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/resnet50_thrift/)。
+torchpipe是为工业界所准备的一个独立作用于底层加速库（如tensorrt，opencv，torchscript）以及 RPC（如thrift, gRPC）之间的多实例流水线并行库，助力使用者能在部署阶段节约更多的硬件资源，帮助产品应用落地。此教程主要针对初级用户，即对于加速相关的理论知识处于入门阶段，具有一定的 Python基础，能够阅读简单代码的用户。此内容主要包括使用torchpipe进行服务部署加速的使用方法、性能和效果差异对比等。本文档的完整代码见可详见[resnet50_thrift](https://github.com/torchpipe/torchpipe/blob/develop/examples/resnet50_thrift/)。
 
 
 
@@ -89,7 +89,7 @@ self.classification_engine = torch2trt(resnet50, [input_shape],
 
 
 
-整体的线上服务部署代码见[main_trt.py](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/resnet50_thrift/main_trt.py)
+整体的线上服务部署代码见[main_trt.py](https://github.com/torchpipe/torchpipe/blob/develop/examples/resnet50_thrift/main_trt.py)
 
 :::tip
 因为TensorRT不是线程安全的，所以利用这种方法进行模型加速时，服务部署过程中需要加锁（`with self.lock:`）处理。
@@ -108,7 +108,7 @@ self.classification_engine = torch2trt(resnet50, [input_shape],
 
 ![](images/quick_start_new_user/torchpipe.png)
 
-利用torchpipe对本服务部署进行调整，整体的线上服务部署代码见[main_torchpipe.py](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/resnet50_thrift/main_torchpipe.py),核心函数调整如下：
+利用torchpipe对本服务部署进行调整，整体的线上服务部署代码见[main_torchpipe.py](https://github.com/torchpipe/torchpipe/blob/develop/examples/resnet50_thrift/main_torchpipe.py),核心函数调整如下：
 
 ```py
 # ------- main -------
@@ -210,7 +210,7 @@ std="58.395, 57.120, 57.375" # 255*"0.229, 0.224, 0.225"
 ## 4 性能和效果对比
 `python test_tools.py --img_dir /your/testimg/path/ --port 8095 --request_client 10 --request_batch 1
 `
-测试具体代码见[client_qps.py](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/resnet50_thrift/client_qps.py)
+测试具体代码见[client_qps.py](https://github.com/torchpipe/torchpipe/blob/develop/examples/resnet50_thrift/client_qps.py)
 
 采用相同的thrift的服务接口，测试机器3080,cpu 36核, 并发数10
 

i18n/zh/docusaurus-plugin-content-docs/current/tools/quantization.mdx

@@ -31,7 +31,7 @@ tensorrt 的训练后量化过程主要包含两步：
 
 训练时量化除了需要正常训练的模型提供预训练参数，也需要训练后量化（ptq）提供量化的预训练参数。
 
-我们集成了[calib_tools](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/calib_tools.py)，可做参考.
+我们集成了[calib_tools](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/calib_tools.py)，可做参考.
 
 - 定义calibrater：
 ```python
@@ -99,17 +99,17 @@ calib_tools.save_onnx(q_model, f"model_name_qat.onnx")
 #### 不改变backbone，直接量化
 仿照官方示例，我们对resnet进行了分步骤实验：
 
-- 下载训练数据：[代码](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/download_data.py)
-- 训练10个epoch获得resnet50模型：[代码](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/fp32_train.py), 精度98.44%
-- (optinal)pytorch ptq：[代码](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/ptq.py), 精度96.64%（max）
-- (optinal)pytorch qat：[代码](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/qat.py), 精度98.26%. 
+- 下载训练数据：[代码](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/download_data.py)
+- 训练10个epoch获得resnet50模型：[代码](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/fp32_train.py), 精度98.44%
+- (optinal)pytorch ptq：[代码](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/ptq.py), 精度96.64%（max）
+- (optinal)pytorch qat：[代码](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/qat.py), 精度98.26%. 
 
 #### mse + 残差融合 {#mseadd}
 
 以上resnet的训练，采用max方式量化，并且没有对Add进行融合，导致tensorrt运行速度未达预期。以下将Add在int8下进行融合并换用mse模式后的结果：
 
-- ptq：[代码](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/ptq_merge_residual.py), 精度94.34%（mse）
-- qat：[代码](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/qat_merge_residual.py), 精度95.82%
+- ptq：[代码](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/ptq_merge_residual.py), 精度94.34%（mse）
+- qat：[代码](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/qat_merge_residual.py), 精度95.82%
 
 #### pytorch下结果汇总
 | Model                    | Accuracy |  Performance |备注 |
@@ -122,9 +122,9 @@ calib_tools.save_onnx(q_model, f"model_name_qat.onnx")
 ### torchpipe下测试结果汇总
 以下使用torchpipe加载生成的onnx进行测试：
 
-- 导出onnx：[代码](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/export_onnx_merge_residual.py)
-- 使用torchpipe加载fp32-onnx并进行ptq： [代码](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/torchpipe_ptq_test.py)
-- 使用torchpipe加载qat-onnx进行测试： [代码](https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/torchpipe_qat_test.py)
+- 导出onnx：[代码](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/export_onnx_merge_residual.py)
+- 使用torchpipe加载fp32-onnx并进行ptq： [代码](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/torchpipe_ptq_test.py)
+- 使用torchpipe加载qat-onnx进行测试： [代码](https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/torchpipe_qat_test.py)
 
 
 | Model                    | Accuracy |  Performance |备注 |
@@ -133,4 +133,4 @@ calib_tools.save_onnx(q_model, f"model_name_qat.onnx")
 | tensorrt's native int8 |98.26%   | -  | |
 | qat   | 98.67%   | -    |  [onnxruntime下精度]为98.69%。   |
 
-[onnxruntime下精度]: https://github.com/torchpipe/torchpipe/-/blob/develop/examples/int8/qat/onnxruntime_qat_test.py
+[onnxruntime下精度]: https://github.com/torchpipe/torchpipe/blob/develop/examples/int8/qat/onnxruntime_qat_test.py