delldu

Fix Jittor Utils Config Bugs

1

## 1. Two Bugs for 'Jittor Utils Config': 1) **Bug 1**: Command option for 'python -m jittor_utils.config' ```$ python -m jittor_utils.config``` Usage: /home/dell/miniconda3/envs/numba/bin/python -m jittor_utils.config --include-flags|**--link-flags**|--cxx-flags|--cxx-example|--help **Expected:** Usage: /home/dell/miniconda3/envs/numba/bin/python -m...

请求Deploy Jittor Module using C++

1

TVM 用户有"Deploy TVM Module using C++"， [请参阅 https://tvm.apache.org/docs/how_to/deploy/cpp_deploy.html](https://tvm.apache.org/docs/how_to/deploy/cpp_deploy.html)， 同样Jittor用户希望有"Deploy **Jittor** Module using C++"。 ### 1.非常、非常、非常重要 1）我们投入时间和精力去训练模型，唯一的目的就是应用（推理），没有它，训练模型有什么用？还不如训练小狗小猫有意思，对吗？ 2）如果推理还要Jittor运行环境，就象普通人背个千斤大石头跑步，令人崩溃，更重要的是，大多数人背不动，死路一条。 ### 2.最高优先级别的任务 1）我们理解Jittor开发有优先级，前期可能是验证**可行**，这个已经实现，**非常不错**！但现在要验证**有用**，所以请求是现在最高优先级的任务，**因为它是唯一让模型有价值的任务**。 2）这件事对核心开发团队来讲，不难，工程量不大，因为理论上，模型能跑起来了！前端虽是python, 但真正动起来的是机器码，形式上表现为c/c++/cuda等生成的静动态库，C/C++可无缝使用它们，所以现在的任务就是收集Jittor运行数据，把它们**打包成库**（libx.a/libx.so之类)供给用户调用，就像用线把珍珠串起来，**Jittor缺这根金丝线！** 3）粗略的流程是：Create Jittor Model --> Trainning --> Test OK...

Support CPU Test.

1

Fix Bug: model load wrong parameters and forward with wrong data type when CPU test.

Request Help: 1) Environment 2) Examples

13

Dear friend, We are interested in your work, but found it is hard to running source code. The key problems are: 1) What's your work environment: python version, sklearn version,...

1. Add file convert.sh, train.sh, test.sh for end user reference. 2. Add command options for large, xlarge model config. 3. Clear guide for README.

BUG: dcn_v2_cpu_forward

1

Dear friend, I check function dcn_v2_cpu_forward, think the following code is wrong: auto ones_T = at::transpose(ones.contiguous(), 2, 0); ones_T = at::mul(ones_T, bias.contiguous()); ones_T = at::transpose(ones_T, 2, 0); output_n = at::add(output_n,...

Question About Super Resolution

1

Dear friend, First all, we hope express our thanks for your hard work. Second, we have an question: Why super resolution output size is same as input images ? Best...

Dear friend, Thanks for your good work at first. We found out of CUDA memory issue with 2080ti when running command "python test_inference.py", so we make a little change for...

As title.

**Good project** need good makefile, please accept it, thanks.