merge master #13
Merged
merge master #13
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
Co-authored-by: oneflow-ci-bot <[email protected]>
fix the error: AttributeError: 'UNet2DConditionModel' object has no attribute 'cpg'
…nal/cuda.cmake to avoid llvm15 link error (#10373)
Co-authored-by: oneflow-ci-bot <[email protected]>
文档：  --------- Co-authored-by: oneflow-ci-bot <[email protected]>
adaptation of huawei ascend910b chip on oneflow: - Oneflow-Inc/OneTeam#2181 --------- Co-authored-by: Shenghang Tsai <[email protected]>
Co-authored-by: oneflow-ci-bot <[email protected]>
本Pr实现了OneFlow Insight模块，相关issue：Oneflow-Inc/OneTeam#2162 当我们需要profiling cuda kernel执行时间/瓶颈分析时，通常会基于nvidia提供的nsys指令，生成对应的profile文件（早期的.qdrep以及现在的.nsys-rep）并用Nvidia的GUI软件Nsight Systems来可视化分析、查看。 在nsys生成profile文件的同时，还会生成平台无关的数据信息，记录在.sqlite文件中，OneFlow Insight模块就可以通过解析.sqlite，来生成符合Google Chrome Trace Event格式的JSON文件，使得可以直接通过Chrome或者Edge浏览器，通过`chrome://tracing/` 或 `edge://tracing/`来解析和渲染此JSON文件，从而进行可视化分析、查看，效果如下： <img width="1320" alt="image" src="https://github.com/Oneflow-Inc/oneflow/assets/28823622/cbfab9bc-47bd-474c-8f39-e145348db17d"> --------- Co-authored-by: oneflow-ci-bot <[email protected]>
Co-authored-by: Houjiang Chen <[email protected]>
python3 -c "import timm" 输出如下： ``` /data/home/wangyi/workspace/oneflow-public/python/oneflow/jit/__init__.py:19: UserWarning: The oneflow.jit interface is just to align the torch.jit interface and has no practical significance. warnings.warn( /data/home/wangyi/workspace/oneflow-public/python/oneflow/jit/__init__.py:134: UserWarning: The oneflow.jit.Final interface is just to align the torch.jit.Final interface and has no practical significance. warnings.warn( /data/home/wangyi/workspace/oneflow-public/python/oneflow/jit/__init__.py:31: UserWarning: The oneflow.jit.script interface is just to align the torch.jit.script interface and has no practical significance. warnings.warn( /home/wangyi/miniconda3/envs/py10/lib/python3.10/site-packages/torchvision/io/image.py:13: UserWarning: Failed to load image Python extension: 'load_library is not implemented'If you don't plan on using image functionality from `torchvision.io`, you can ignore this warning. Otherwise, there might be something wrong with your environment. Did you have `libjpeg` or `libpng` installed before building `torchvision` from source? warn( /data/home/wangyi/workspace/oneflow-public/python/oneflow/jit/__init__.py:49: UserWarning: The oneflow.jit.unused interface is just to align the torch.jit.unused interface and has no practical significance. warnings.warn( /data/home/wangyi/workspace/oneflow-public/python/oneflow/onnx/symbolic_helper.py:21: UserWarning: The oneflow.onnx.parse_args interface is just to align the torch.onnx.parse_args interface and has no practical significance. warnings.warn( /data/home/wangyi/workspace/oneflow-public/python/oneflow/jit/__init__.py:57: UserWarning: The oneflow.jit._script_if_tracing interface is just to align the torch.jit._script_if_tracing interface and has no practical significance. warnings.warn( /data/home/wangyi/workspace/oneflow-public/python/oneflow/onnx/__init__.py:26: UserWarning: The oneflow.onnx.register_custom_op_symbolic interface is just to align the torch.onnx.register_custom_op_symbolic interface and has no practical significance. warnings.warn( /data/home/wangyi/workspace/oneflow-public/python/oneflow/jit/__init__.py:65: UserWarning: The oneflow.jit._overload_method interface is just to align the torch.jit._overload_method interface and has no practical significance. warnings.warn( /data/home/wangyi/workspace/oneflow-public/python/oneflow/jit/__init__.py:134: UserWarning: The oneflow.jit.Final interface is just to align the torch.jit.Final interface and has no practical significance. warnings.warn( /data/home/wangyi/workspace/oneflow-public/python/oneflow/jit/__init__.py:31: UserWarning: The oneflow.jit.script interface is just to align the torch.jit.script interface and has no practical significance. warnings.warn( /data/home/wangyi/workspace/oneflow-public/python/oneflow/jit/__init__.py:38: UserWarning: The oneflow.jit.ignore interface is just to align the torch.jit.ignore interface and has no practical significance. warnings.warn( /data/home/wangyi/workspace/oneflow-public/python/oneflow/jit/__init__.py:140: UserWarning: The oneflow.jit.interface interface is just to align the torch.jit.interface interface and has no practical significance. warnings.warn( ```
* register mem_get_info * add unittest * refine unittest * Update env.cpp * format * format * format * format * rename GetMemInfo to CudaGetMemoInfo
重构 mock_torch 模块
close #10397 这里只需要在参数检查的时候对 is_grads_batched 做处理就行了，不需要侵入到 AutogradEgnine 里。 实际后向计算的时候，会自己做 broadcast 操作，如果计算错误，是算子对 broadcast 支持的不全。 --------- Co-authored-by: wyg1997 <[email protected]>
Oneflow-Inc/oneflow#10395 这个 PR 里多此一举给 jit/\_\_init__.py 加了一个 warning，导致 import oneflow 的时候就会显示这个 warning，这里去掉
fix #10405 Co-authored-by: wyg1997 <[email protected]>
…_compiler.compile_from_torch (#10408) resubmit for PR #10404 --------- Co-authored-by: Yinggang Wang <[email protected]>
Before this commit, error messge is: **inconsistent tensor size, expected all tensor to have the same number of elements, but got 640000 and 742400** After this commit, error message is **inconsistent tensor size, expected all tensor to have the same number of elements, but got (1,256,50,50) and (1,256,50,58)**
…of the host compiler. (#10415)
## 分支工作 添加jvp、jacobian、hessian、hvp、vhp接口到funcitional.py文件，添加测试代码到test_autograd_functional.py文件 ## 代码修改部分 1. \_construct_standard_basis_for(): 因fill_ 不支持 NonContiguous，将fill_赋值替换为循环赋值。 (Oneflow-Inc/oneflow#10394) 2. \_construct_standard_basis_for(): 函数参数tensors: Tuple[torch.Tensor, ...]的写法会导致循环导入torch，因此将后面的描述去掉，变为tensors 3. _grad_preprocess:因oneflow.torch没有is_sparse属性，只采用res.append(inp.view_as(inp))方式创建新tensor 4. _jacfwd: 删除vectorize情况的代码 5. 在test_autograd_functional.py文件中添加jvp、jacobian、hessian、hvp、vhp测试代码 ## 添加接口  --------- Co-authored-by: Wang Yi <[email protected]> Co-authored-by: oneflow-ci-bot <[email protected]>
Co-authored-by: oneflow-ci-bot <[email protected]>
可以减少一点点编译时间
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in 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.
merge master