pytorch - 如何将 torch.utils.tensorboard 的 SummaryWriter add

pytorch - 如何将 torch.utils.tensorboard 的 SummaryWriter add_graph 与字典输出一起使用？

转载作者：行者123 更新时间：2023-12-05 07:15:12

我有一个继承自 nn.Module 的 PyTorch 模型并且有一个 forward返回包含多个张量(存储为值)的字典的方法。当我尝试使用 torch.utils.tensorboard 将此模型记录到 TensorBoard 时的 SummaryWriter.add_graph()方法我收到以下错误:

RuntimeError("Tracer cannot infer type of <output dictionary>

如何使用 add_graph()当模型的 forward()方法返回字典？

编辑 1:完成错误堆栈跟踪:

WARNING:root:This caffe2 python run does not have GPU support. Will run in CPU only mode.
Only tensors or tuples of tensors can be output from traced functions (getOutput at /pytorch/torch/csrc/jit/tracer.cpp:208)
frame #0: c10::Error::Error(c10::SourceLocation, std::string const&) + 0x33 (0x7fc661bd4273 in /usr/local/lib/python3.6/dist-packages/torch/lib/libc10.so)
frame #1: torch::jit::tracer::TracingState::getOutput(c10::IValue const&) + 0x3e8 (0x7fc664cf77b8 in /usr/local/lib/python3.6/dist-packages/torch/lib/libtorch.so)
frame #2: torch::jit::tracer::exit(std::vector<c10::IValue, std::allocator<c10::IValue> > const&) + 0x3c (0x7fc664cf7aec in /usr/local/lib/python3.6/dist-packages/torch/lib/libtorch.so)
frame #3: <unknown function> + 0x47b0f2 (0x7fc66e4820f2 in /usr/local/lib/python3.6/dist-packages/torch/lib/libtorch_python.so)
frame #4: <unknown function> + 0x48f061 (0x7fc66e496061 in /usr/local/lib/python3.6/dist-packages/torch/lib/libtorch_python.so)
frame #5: <unknown function> + 0x19da64 (0x7fc66e1a4a64 in /usr/local/lib/python3.6/dist-packages/torch/lib/libtorch_python.so)
frame #6: _PyCFunction_FastCallDict + 0x35c (0x5674fc in /usr/bin/python3.6)
frame #7: /usr/bin/python3.6() [0x50abb3]
frame #8: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #9: <unknown function> + 0x18d10 (0x7fc6965ded10 in /home/rylan/.PyCharmCE2019.3/system/cythonExtensions/_pydevd_frame_eval_ext/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so)
frame #10: /usr/bin/python3.6() [0x508245]
frame #11: /usr/bin/python3.6() [0x50a080]
frame #12: /usr/bin/python3.6() [0x50aa7d]
frame #13: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #14: <unknown function> + 0x18d10 (0x7fc6965ded10 in /home/rylan/.PyCharmCE2019.3/system/cythonExtensions/_pydevd_frame_eval_ext/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so)
frame #15: /usr/bin/python3.6() [0x508245]
frame #16: /usr/bin/python3.6() [0x50a080]
frame #17: /usr/bin/python3.6() [0x50aa7d]
frame #18: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #19: <unknown function> + 0x18d10 (0x7fc6965ded10 in /home/rylan/.PyCharmCE2019.3/system/cythonExtensions/_pydevd_frame_eval_ext/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so)
frame #20: /usr/bin/python3.6() [0x508245]
frame #21: /usr/bin/python3.6() [0x50a080]
frame #22: /usr/bin/python3.6() [0x50aa7d]
frame #23: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #24: <unknown function> + 0x18d10 (0x7fc6965ded10 in /home/rylan/.PyCharmCE2019.3/system/cythonExtensions/_pydevd_frame_eval_ext/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so)
frame #25: /usr/bin/python3.6() [0x508245]
frame #26: /usr/bin/python3.6() [0x50a080]
frame #27: /usr/bin/python3.6() [0x50aa7d]
frame #28: _PyEval_EvalFrameDefault + 0x1220 (0x50d390 in /usr/bin/python3.6)
frame #29: <unknown function> + 0x18d10 (0x7fc6965ded10 in /home/rylan/.PyCharmCE2019.3/system/cythonExtensions/_pydevd_frame_eval_ext/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so)
frame #30: /usr/bin/python3.6() [0x508245]
frame #31: _PyFunction_FastCallDict + 0x2e2 (0x509642 in /usr/bin/python3.6)
frame #32: /usr/bin/python3.6() [0x595311]
frame #33: /usr/bin/python3.6() [0x54a6ff]
frame #34: /usr/bin/python3.6() [0x551b81]
frame #35: _PyObject_FastCallKeywords + 0x19c (0x5aa6ec in /usr/bin/python3.6)
frame #36: /usr/bin/python3.6() [0x50abb3]
frame #37: _PyEval_EvalFrameDefault + 0x1220 (0x50d390 in /usr/bin/python3.6)
frame #38: <unknown function> + 0x18d10 (0x7fc6965ded10 in /home/rylan/.PyCharmCE2019.3/system/cythonExtensions/_pydevd_frame_eval_ext/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so)
frame #39: /usr/bin/python3.6() [0x509d48]
frame #40: /usr/bin/python3.6() [0x50aa7d]
frame #41: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #42: <unknown function> + 0x18d10 (0x7fc6965ded10 in /home/rylan/.PyCharmCE2019.3/system/cythonExtensions/_pydevd_frame_eval_ext/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so)
frame #43: /usr/bin/python3.6() [0x508245]
frame #44: /usr/bin/python3.6() [0x516915]
frame #45: /usr/bin/python3.6() [0x50a8af]
frame #46: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #47: <unknown function> + 0x18d10 (0x7fc6965ded10 in /home/rylan/.PyCharmCE2019.3/system/cythonExtensions/_pydevd_frame_eval_ext/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so)
frame #48: /usr/bin/python3.6() [0x508245]
frame #49: /usr/bin/python3.6() [0x50a080]
frame #50: /usr/bin/python3.6() [0x50aa7d]
frame #51: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #52: <unknown function> + 0x18d10 (0x7fc6965ded10 in /home/rylan/.PyCharmCE2019.3/system/cythonExtensions/_pydevd_frame_eval_ext/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so)
frame #53: /usr/bin/python3.6() [0x509d48]
frame #54: /usr/bin/python3.6() [0x50aa7d]
frame #55: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #56: /usr/bin/python3.6() [0x508245]
frame #57: /usr/bin/python3.6() [0x50a080]
frame #58: /usr/bin/python3.6() [0x50aa7d]
frame #59: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #60: /usr/bin/python3.6() [0x509d48]
frame #61: /usr/bin/python3.6() [0x50aa7d]
frame #62: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #63: /usr/bin/python3.6() [0x508245]

Error occurs, No graph saved
Traceback (most recent call last):
  File "/usr/lib/python3.6/contextlib.py", line 99, in __exit__
    self.gen.throw(type, value, traceback)
  File "/usr/local/lib/python3.6/dist-packages/torch/onnx/utils.py", line 42, in set_training
    yield
  File "/usr/local/lib/python3.6/dist-packages/torch/utils/tensorboard/_pytorch_graph.py", line 234, in graph
    raise e
  File "/usr/local/lib/python3.6/dist-packages/torch/utils/tensorboard/_pytorch_graph.py", line 229, in graph
    trace = torch.jit.trace(model, args)
  File "/usr/local/lib/python3.6/dist-packages/torch/jit/__init__.py", line 772, in trace
    check_tolerance, _force_outplace, _module_class)
  File "/usr/local/lib/python3.6/dist-packages/torch/jit/__init__.py", line 904, in trace_module
    module._c._create_method_from_trace(method_name, func, example_inputs, var_lookup_fn, _force_outplace)
RuntimeError: Only tensors or tuples of tensors can be output from traced functions (getOutput at /pytorch/torch/csrc/jit/tracer.cpp:208)
frame #0: c10::Error::Error(c10::SourceLocation, std::string const&) + 0x33 (0x7fc661bd4273 in /usr/local/lib/python3.6/dist-packages/torch/lib/libc10.so)
frame #1: torch::jit::tracer::TracingState::getOutput(c10::IValue const&) + 0x3e8 (0x7fc664cf77b8 in /usr/local/lib/python3.6/dist-packages/torch/lib/libtorch.so)
frame #2: torch::jit::tracer::exit(std::vector<c10::IValue, std::allocator<c10::IValue> > const&) + 0x3c (0x7fc664cf7aec in /usr/local/lib/python3.6/dist-packages/torch/lib/libtorch.so)
frame #3: <unknown function> + 0x47b0f2 (0x7fc66e4820f2 in /usr/local/lib/python3.6/dist-packages/torch/lib/libtorch_python.so)
frame #4: <unknown function> + 0x48f061 (0x7fc66e496061 in /usr/local/lib/python3.6/dist-packages/torch/lib/libtorch_python.so)
frame #5: <unknown function> + 0x19da64 (0x7fc66e1a4a64 in /usr/local/lib/python3.6/dist-packages/torch/lib/libtorch_python.so)
frame #6: _PyCFunction_FastCallDict + 0x35c (0x5674fc in /usr/bin/python3.6)
frame #7: /usr/bin/python3.6() [0x50abb3]
frame #8: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #9: <unknown function> + 0x18d10 (0x7fc6965ded10 in /home/rylan/.PyCharmCE2019.3/system/cythonExtensions/_pydevd_frame_eval_ext/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so)
frame #10: /usr/bin/python3.6() [0x508245]
frame #11: /usr/bin/python3.6() [0x50a080]
frame #12: /usr/bin/python3.6() [0x50aa7d]
frame #13: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #14: <unknown function> + 0x18d10 (0x7fc6965ded10 in /home/rylan/.PyCharmCE2019.3/system/cythonExtensions/_pydevd_frame_eval_ext/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so)
frame #15: /usr/bin/python3.6() [0x508245]
frame #16: /usr/bin/python3.6() [0x50a080]
frame #17: /usr/bin/python3.6() [0x50aa7d]
frame #18: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #19: <unknown function> + 0x18d10 (0x7fc6965ded10 in /home/rylan/.PyCharmCE2019.3/system/cythonExtensions/_pydevd_frame_eval_ext/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so)
frame #20: /usr/bin/python3.6() [0x508245]
frame #21: /usr/bin/python3.6() [0x50a080]
frame #22: /usr/bin/python3.6() [0x50aa7d]
frame #23: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #24: <unknown function> + 0x18d10 (0x7fc6965ded10 in /home/rylan/.PyCharmCE2019.3/system/cythonExtensions/_pydevd_frame_eval_ext/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so)
frame #25: /usr/bin/python3.6() [0x508245]
frame #26: /usr/bin/python3.6() [0x50a080]
frame #27: /usr/bin/python3.6() [0x50aa7d]
frame #28: _PyEval_EvalFrameDefault + 0x1220 (0x50d390 in /usr/bin/python3.6)
frame #29: <unknown function> + 0x18d10 (0x7fc6965ded10 in /home/rylan/.PyCharmCE2019.3/system/cythonExtensions/_pydevd_frame_eval_ext/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so)
frame #30: /usr/bin/python3.6() [0x508245]
frame #31: _PyFunction_FastCallDict + 0x2e2 (0x509642 in /usr/bin/python3.6)
frame #32: /usr/bin/python3.6() [0x595311]
frame #33: /usr/bin/python3.6() [0x54a6ff]
frame #34: /usr/bin/python3.6() [0x551b81]
frame #35: _PyObject_FastCallKeywords + 0x19c (0x5aa6ec in /usr/bin/python3.6)
frame #36: /usr/bin/python3.6() [0x50abb3]
frame #37: _PyEval_EvalFrameDefault + 0x1220 (0x50d390 in /usr/bin/python3.6)
frame #38: <unknown function> + 0x18d10 (0x7fc6965ded10 in /home/rylan/.PyCharmCE2019.3/system/cythonExtensions/_pydevd_frame_eval_ext/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so)
frame #39: /usr/bin/python3.6() [0x509d48]
frame #40: /usr/bin/python3.6() [0x50aa7d]
frame #41: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #42: <unknown function> + 0x18d10 (0x7fc6965ded10 in /home/rylan/.PyCharmCE2019.3/system/cythonExtensions/_pydevd_frame_eval_ext/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so)
frame #43: /usr/bin/python3.6() [0x508245]
frame #44: /usr/bin/python3.6() [0x516915]
frame #45: /usr/bin/python3.6() [0x50a8af]
frame #46: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #47: <unknown function> + 0x18d10 (0x7fc6965ded10 in /home/rylan/.PyCharmCE2019.3/system/cythonExtensions/_pydevd_frame_eval_ext/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so)
frame #48: /usr/bin/python3.6() [0x508245]
frame #49: /usr/bin/python3.6() [0x50a080]
frame #50: /usr/bin/python3.6() [0x50aa7d]
frame #51: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #52: <unknown function> + 0x18d10 (0x7fc6965ded10 in /home/rylan/.PyCharmCE2019.3/system/cythonExtensions/_pydevd_frame_eval_ext/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so)
frame #53: /usr/bin/python3.6() [0x509d48]
frame #54: /usr/bin/python3.6() [0x50aa7d]
frame #55: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #56: /usr/bin/python3.6() [0x508245]
frame #57: /usr/bin/python3.6() [0x50a080]
frame #58: /usr/bin/python3.6() [0x50aa7d]
frame #59: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #60: /usr/bin/python3.6() [0x509d48]
frame #61: /usr/bin/python3.6() [0x50aa7d]
frame #62: _PyEval_EvalFrameDefault + 0x449 (0x50c5b9 in /usr/bin/python3.6)
frame #63: /usr/bin/python3.6() [0x508245]

最佳答案

你可以使用这个模型包装器:

from collections import namedtuple

从键入 import Any

类 ModelWrapper(torch.nn.Module):"""具有 dict/list 右值的模型的包装器类。"""

def __init__(self, model: torch.nn.Module) -> None:
    """
    Init call.
    """
    super().__init__()
    self.model = model

def forward(self, input_x: torch.Tensor) -> Any:
    """
    Wrap forward call.
    """
    data = self.model(input_x)

    if isinstance(data, dict):
        data_named_tuple = namedtuple("ModelEndpoints", sorted(data.keys()))  # type: ignore
        data = data_named_tuple(**data)  # type: ignore

    elif isinstance(data, list):
        data = tuple(data)

    return data

关于pytorch - 如何将 torch.utils.tensorboard 的 SummaryWriter add_graph 与字典输出一起使用？，我们在Stack Overflow上找到一个类似的问题： https://stackoverflow.com/questions/59656306/

文章推荐： node.js - Gatsby 的环境变量 "env.cmd not found"

文章推荐： tensorflow - 如何从 TF Hub 获取 Bert tokenizer 的 vocab 文件

文章推荐： Azure应用服务容器不断重新启动

文章推荐： c# - 如何使用c#中的程序检查堆栈内存的大小

pytorch - PyTorch 为什么叫 PyTorch？
已关闭。此问题不符合Stack Overflow guidelines 。目前不接受答案。这个问题似乎与 help center 中定义的范围内的编程无关。 . 已关闭 3 年前。此帖子于去年编辑
pytorch - 验证阶段完成后gpu内存仍然被占用，pytorch
据我所知，在使用 GPU 训练和验证模型时，GPU 内存主要用于加载数据，向前和向后。据我所知，我认为 GPU 内存使用应该相同 1) 训练前，2) 训练后，3) 验证前，4) 验证后。但在我的例子中
pytorch - PyTorch 中复数的矩阵乘法
我正在尝试在 PyTorch 中将两个复数矩阵相乘，看起来 the torch.matmul functions is not added yet to PyTorch library for com
pytorch - Pytorch 中软标签的交叉熵
我正在尝试定义二分类问题的损失函数。但是，目标标签不是硬标签0，1，而是0~1之间的一个 float 。 Pytorch 中的 torch.nn.CrossEntropy 不支持软标签，所以我想自己写
pytorch - PyTorch 数据集应该返回什么？
我正在尝试让 PyTorch 与 DataLoader 一起工作，据说这是处理小批量的最简单方法，在某些情况下这是获得最佳性能所必需的。 DataLoader 需要一个数据集作为输入。大多数关于 D
pytorch - Pytorch DataLoader迭代顺序是否稳定？
Pytorch Dataloader 的迭代顺序是否保证相同(在温和条件下)？例如: dataloader = DataLoader(my_dataset, batch_size=4,
pytorch - Pytorch NLLLOSS的理解
PyTorch 的负对数似然损失，nn.NLLLoss定义为: 因此，如果以单批处理的标准重量计算损失，则损失的公式始终为: -1 * (prediction of model for correct
pytorch - PyTorch:new_ones与1
在PyTorch中，new_ones()与ones()有什么区别。例如， x2.new_ones(3,2, dtype=torch.double) 与 torch.ones(3,2, dtype=to
pytorch - PyTorch 中复杂掩码的最大池化
假设我有一个矩阵 src带形状(5, 3)和一个 bool 矩阵 adj带形状(5, 5)如下， src = tensor([[ 0, 1, 2], [ 3, 4,
pytorch - PyTorch 如何在张量的每一行中随机设置固定数量的元素
我想知道如果不在第 4 行中使用“for”循环，下面的代码是否有更有效的替代方案？ import torch n, d = 37700, 7842 k = 4 sample = torch.cat([
pytorch - PyTorch 中的自定义损失函数
我有三个简单的问题。如果我的自定义损失函数不可微会发生什么？ pytorch 会通过错误还是做其他事情？如果我在我的自定义函数中声明了一个损失变量来表示模型的最终损失，我应该放 requires_
pytorch - PyTorch 中参数与张量的区别
我想知道 PyTorch Parameter 和 Tensor 的区别？现有answer适用于使用变量的旧 PyTorch？最佳答案这就是 Parameter 的全部想法。类(附加)在单个图像中
pytorch - Pytorch 中是否有一种方法可以以可以反向传播的方式计算唯一值的数量？
给定以下张量(这是网络的结果 [注意 grad_fn]): tensor([121., 241., 125., 1., 108., 238., 125., 121., 13., 117., 12
pytorch - Pytorch 线性模块类定义中的常量
什么是__constants__在 pytorch class Linear(Module):定义于 https://pytorch.org/docs/stable/_modules/torch/nn
pytorch - pytorch conv2d的源代码在哪里？
我在哪里可以找到pytorch函数conv2d的源代码？它应该在 torch.nn.functional 中，但我只找到了 _add_docstr 行，如果我搜索conv2d。我在这里看了: ht
pytorch - PyTorch 中的默认膨胀值
如 documentation 中所述在 PyTorch 中，Conv2d 层使用默认膨胀为 1。这是否意味着如果我想创建一个简单的 conv2d 层，我必须编写 nn.conv2d(in_chann
pytorch - PyTorch 如何实现反向卷积？
我阅读了 Pytorch 的源代码，发现它没有实现 convolution_backward 很奇怪。函数，唯一的 convolution_backward_overrideable 函数是直接引发错
pytorch - pytorch 中的一种热门编码
我对编码真的很陌生，现在我正在尝试将我的标签变成一种热门编码。我已经完成将 np.array 传输到张量，如下所示 tensor([4., 4., 4., 4., 4., 4., 4., 4., 4.
pytorch - PyTorch 中用于文本输入的卷积神经网络
我正在尝试实现 text classification model使用CNN。据我所知，对于文本数据，我们应该使用一维卷积。我在 pytorch 中看到了一个使用 Conv2d 的示例，但我想知道如何
pytorch - Pytorch 中类别不平衡的多标签分类
我有一个多标签分类问题，我正试图用 Pytorch 中的 CNN 解决这个问题。我有 80,000 个训练示例和 7900 个类；每个示例可以同时属于多个类，每个示例的平均类数为 130。问题是我的

行者123

个人简介

我是一名优秀的程序员,十分优秀！

作者热门文章

滴滴打车优惠券免费领取

全站热门文章

首页

博学

6Ren·AI

商城

pytorch - 如何将 torch.utils.tensorboard 的 SummaryWriter add_graph 与字典输出一起使用？