non-static forward method will be removed in 1.3

[BingkunZhao]

trafficstars

Hi, the version of my pytorch is 1.3. and the python version is 3.6. when I run the code

y = net(x)

which is for the detection. the warning is arise: UserWarning: Legacy autograd function with non-static forward method is deprecated and will be removed in 1.3. Please use new-style autograd function with static forward method. (Example: https://pytorch.org/docs/stable/autograd.html#torch.autograd.Function)

I am very confused about how to solve that problem, so is there anyone have a clue?

[qcy007]

I have same problem like you. my torch version is 1.4.0. But it seems that the problem has no effect to training. have you solve it?

[kevinchan04]

Degrade the pytorch version to 1.2.0 or below will solve this problem.

[brijhub]

Hi, the version of my pytorch is 1.3. and the python version is 3.6. when I run the code

y = net(x)

which is for the detection. the warning is arise: UserWarning: Legacy autograd function with non-static forward method is deprecated and will be removed in 1.3. Please use new-style autograd function with static forward method. (Example: https://pytorch.org/docs/stable/autograd.html#torch.autograd.Function)

I am very confused about how to solve that problem, so is there anyone have a clue? Did you get the solution ?

[ayukat1016]

Hi, version of colab is PyTorch1.5.0, I modified detection.py and ssd.py. so that it works without warning and error.

Since detection.py have torch.autograd.Function, I delete def init() and add @staticmethod. I also add output = self.detect.apply() in ssd.py.

I share my code and you can find in '# PyTorch1.5.0 support new-style autograd function' in source. https://github.com/miyamotok0105/pytorch_handbook/blob/master/chapter7/layers/functions/detection.py

[fywsjz]

@ayukat1016 hello.I have modified the detection file.I am confused about where to add output = self.detection. apply() in the ssd.py file.

[ayukat1016]

@fywsjz Hello. You can find in line 63, 123. https://github.com/miyamotok0105/pytorch_handbook/blob/master/chapter7/ssd.py

[fywsjz]

@ayukat1016 Thank you very much. It worked

[decoli]

Hi, version of colab is PyTorch1.5.0, I modified detection.py and ssd.py. so that it works without warning and error.

Since detection.py have torch.autograd.Function, I delete def init() and add @staticmethod. I also add output = self.detect.apply() in ssd.py.

I share my code and you can find in '# PyTorch1.5.0 support new-style autograd function' in source. https://github.com/miyamotok0105/pytorch_handbook/blob/master/chapter7/layers/functions/detection.py

うまく動作する様になりました、ありがとうございます！

[decoli]

Hi, version of colab is PyTorch1.5.0, I modified detection.py and ssd.py. so that it works without warning and error.

Since detection.py have torch.autograd.Function, I delete def init() and add @staticmethod. I also add output = self.detect.apply() in ssd.py.

I share my code and you can find in '# PyTorch1.5.0 support new-style autograd function' in source. https://github.com/miyamotok0105/pytorch_handbook/blob/master/chapter7/layers/functions/detection.py

Hi, it can work well like below:

Finished loading model!
im_detect: 1/4952 1.888s
im_detect: 2/4952 0.145s
...

Evaluating detections
Writing aeroplane VOC results file
Writing bicycle VOC results file
...

VOC07 metric? Yes

But after that, it will error: Exception has occurred: KeyError at 317, eval.py R = class_recs[image_ids[d]]

And here:

class_recs
{}

image_ids
['000002', '000002', '000002', '000002', '000002', '000002', '000003']

d
0

Do you know what here happening and can you help me? For the time being, I opened new issue here

[fywsjz]

I have not encountered the problem you described, you can try to find the answer in the issues at github, sorry for not being able to help you

| | 封亚炜 | | [email protected] | 签名由网易邮箱大师定制 On 5/28/2020 16:43，Shirui Zhang[email protected] wrote：

Hi, version of colab is PyTorch1.5.0, I modified detection.py and ssd.py. so that it works without warning and error.

Since detection.py have torch.autograd.Function, I delete def init() and add @staticmethod. I also add output = self.detect.apply() in ssd.py.

I share my code and you can find in '# PyTorch1.5.0 support new-style autograd function' in source. https://github.com/miyamotok0105/pytorch_handbook/blob/master/chapter7/layers/functions/detection.py

Hi, it can work well like below:

Finished loading model! im_detect: 1/4952 1.888s im_detect: 2/4952 0.145s ...

Evaluating detections Writing aeroplane VOC results file Writing bicycle VOC results file ...

VOC07 metric? Yes

But after that, it will error: Exception has occurred: KeyError at 317, eval.py R = class_recs[image_ids[d]]

And here:

class_recs {}

image_ids ['000002', '000002', '000002', '000002', '000002', '000002', '000003']

d 0

Do you know what here happening and can you help me?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

[decoli]

I have not encountered the problem you described, you can try to find the answer in the issues at github, sorry for not being able to help you | | 封亚炜 | | [email protected] | 签名由网易邮箱大师定制 On 5/28/2020 16:43，Shirui Zhang[email protected] wrote： Hi, version of colab is PyTorch1.5.0, I modified detection.py and ssd.py. so that it works without warning and error. Since detection.py have torch.autograd.Function, I delete def init() and add @staticmethod. I also add output = self.detect.apply() in ssd.py. I share my code and you can find in '# PyTorch1.5.0 support new-style autograd function' in source. https://github.com/miyamotok0105/pytorch_handbook/blob/master/chapter7/layers/functions/detection.py Hi, it can work well like below: Finished loading model! im_detect: 1/4952 1.888s im_detect: 2/4952 0.145s ... Evaluating detections Writing aeroplane VOC results file Writing bicycle VOC results file ... VOC07 metric? Yes But after that, it will error: Exception has occurred: KeyError at 317, eval.py R = class_recs[image_ids[d]] And here: class_recs {} image_ids ['000002', '000002', '000002', '000002', '000002', '000002', '000003'] d 0 Do you know what here happening and can you help me? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

Thanks anyway.

[decoli]

Hi, version of colab is PyTorch1.5.0, I modified detection.py and ssd.py. so that it works without warning and error.

Since detection.py have torch.autograd.Function, I delete def init() and add @staticmethod. I also add output = self.detect.apply() in ssd.py.

I share my code and you can find in '# PyTorch1.5.0 support new-style autograd function' in source. https://github.com/miyamotok0105/pytorch_handbook/blob/master/chapter7/layers/functions/detection.py

Hi, I modified detection.py and ssd.py as yours.

Then I tried running eval.py, and actually got this problem:

RuntimeError: CUDA out of memory. Tried to allocate 4.08 GiB (GPU 0; 8.00 GiB total capacity; 200.34 MiB already allocated; 5.04 GiB free; 958.00 MiB reserved in total by PyTorch)

The GPU on the machine is 2070 super. In my opinion, its performance should be sufficient.

Then also I tried running eval.py by --cuda False, and it can run: ... im_detect: 997/4952 0.320s im_detect: 998/4952 0.343s ...

Why can't it run on GPU? Do you have any ideas?

---

Win 10 Python 3.6.8 Pytorch 1.5.1 Cuda V11.0.167

[Badaniao]

@fywsjz Hello. You can find in line 63, 123. https://github.com/miyamotok0105/pytorch_handbook/blob/master/chapter7/ssd.py

Hi, version of colab is PyTorch1.5.0, I modified detection.py and ssd.py. so that it works without warning and error. Since detection.py have torch.autograd.Function, I delete def init() and add @staticmethod. I also add output = self.detect.apply() in ssd.py. I share my code and you can find in '# PyTorch1.5.0 support new-style autograd function' in source. https://github.com/miyamotok0105/pytorch_handbook/blob/master/chapter7/layers/functions/detection.py

うまく動作する様になりました、ありがとうございます！

thank you so much for solving this error!!

[actruce]

Hi, version of colab is PyTorch1.5.0, I modified detection.py and ssd.py. so that it works without warning and error.

Since detection.py have torch.autograd.Function, I delete def init() and add @staticmethod. I also add output = self.detect.apply() in ssd.py.

I share my code and you can find in '# PyTorch1.5.0 support new-style autograd function' in source. https://github.com/miyamotok0105/pytorch_handbook/blob/master/chapter7/layers/functions/detection.py

Thank you so much!

[Howietzh]

I got a easy way out: just mordify the base class of Detect to 'object', and then modify the code in ssd.py: if self.phase == "test": with torch.no_grad(): output = self.detect.forward( loc.view(loc.size(0), -1, 4), # loc preds self.softmax(conf.view(conf.size(0), -1, self.num_classes)), # conf preds self.priors.type(type(x.data)) # default boxes )

[cotyyang]

@fywsjz Hello. You can find in line 63, 123. https://github.com/miyamotok0105/pytorch_handbook/blob/master/chapter7/ssd.py

thanks,i solve my issue

[bruce1408]

Hi, version of colab is PyTorch1.5.0, I modified detection.py and ssd.py. so that it works without warning and error.

Since detection.py have torch.autograd.Function, I delete def init() and add @staticmethod. I also add output = self.detect.apply() in ssd.py.

I share my code and you can find in '# PyTorch1.5.0 support new-style autograd function' in source. https://github.com/miyamotok0105/pytorch_handbook/blob/master/chapter7/layers/functions/detection.py

I have question, in pytorch1.5.0, must implement backward, but the class Detect() has no backward function,why? RuntimeError: You must implement the backward function for custom autograd.Function.

[tetsu-kikuchi]

FYI: I heard from my colleague that this error was circumvented by changing

self.detect = Detect(num_classes, 0, 200, 0.01, 0.45)

to

self.detect = Detect(num_classes, 0, 200, 0.01, 0.45).forward

(i.e., just attaching .foward) in ssd.py. https://github.com/amdegroot/ssd.pytorch/blob/master/ssd.py#L48