PytorchStreamReader failed reading zip archive: failed finding central directory (no backtrace available)

[kuloud]

🐛 Bug

To Reproduce

Steps to reproduce the behavior:

1. Module.load(modelFile.absolutePath)
   2.PytorchStreamReader failed reading zip archive

Expected behavior

Load the model file correct.

Environment

• PyTorch Android Version : 1.3.1 ( org.pytorch:pytorch_android:1.3.1 )
• Device Info: Mi 8 / MIUI 11.0.4.0 / Qualcomm Dragon 845

Additional context

onet_epoch.pt.zip

    com.facebook.jni.CppException: [enforce fail at inline_container.cc:137] . PytorchStreamReader failed reading zip archive: failed finding central directory
    (no backtrace available)
        at org.pytorch.Module$NativePeer.initHybrid(Native Method)
        at org.pytorch.Module$NativePeer.<init>(Module.java:70)
        at org.pytorch.Module.<init>(Module.java:25)
        at org.pytorch.Module.load(Module.java:21)

cc @suo

[ngimel]

Can you please provide a minimum script demonstrating the error?

[zdevito]

The attached file does not seem to be a valid TorchScript file (i.e. one generated with .save on a ScriptModule object). Can you provide more detail about how it was generated?

[kuloud]

The attached file does not seem to be a valid TorchScript file (i.e. one generated with .save on a ScriptModule object). Can you provide more detail about how it was generated?

this model file comes from
https://github.com/Sierkinhane/mtcnn-pytorch

the script to generate the file :
https://github.com/Sierkinhane/mtcnn-pytorch/blob/master/mtcnn/train_net/train.py

Thanks.

[kuloud]

Can you please provide a minimum script demonstrating the error?
line 65 in the file:
https://github.com/didi/AoE/blob/master/Android/third_party/pytorch/src/main/kotlin/com/didi/aoe/runtime/pytorch/PyTorchInterpreter.kt

I try to load the attached model file in this way.

[driazati]

The onet_epoch.pt.zip file you provided looks like it's produced by calling torch.save (maybe here?). TorchScript on mobile can only load a compiled model (one saved with torch.jit.save).

You will need to make your model compatible with TorchScript (either via tracing or scripting, details here), then save that via torch.jit.save.

[kuloud]

The onet_epoch.pt.zip file you provided looks like it's produced by calling torch.save (maybe here?). TorchScript on mobile can only load a compiled model (one saved with torch.jit.save).

You will need to make your model compatible with TorchScript (either via tracing or scripting, details here), then save that via torch.jit.save.

Ok, It works for me, thank you.

[saurabhmalviya25]

@kuloud Can you please share the code with scripting.
I am also facing the same issue but unable to resolve it.

[Coderx7]

I faced this issue and my problem was that, the pretrained model I used, had saved the model wholesomely ( saved it like torch.save({'model': model}) and as you know, this is very bad, since in order to use that model, you need to have the same module/file/dir heirarchy or else it will crash.
I tried to convert this model into torch script, it went fine, but upon using I faced this error.

to solve this, I simply first saved the models state_dict() into a new checkpoint file, loaded from the new checkpoint and then converted my model.
hope it helps you as well

[HeyangQin]

In my case, this error was caused by a corrupted saved file. So I switch to older checkpoints and the problem is gone.

[tejan-rgb]

model_mdetr, postprocessor = torch.hub.load('ashkamath/mdetr:main', 'mdetr_efficientnetB5', pretrained=True, return_postprocessor=True)

I am running this and it is showing me this error:

RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory
Help!

[bekirbakar]

In my case, this error was caused by a corrupted saved file. So I switch to older checkpoints and the problem is gone.

I had the same problem today. In my case, it was because of corrupted download files. Somehow it stops downloading and goes to the next process which is unzipping. There the problem occurs.

There must be hash control or something else for downloads and I think an exception should be raised.

[naveenjafer]

@bekirbakar Did you manage to fix this? I am running into the same issue.

[bekirbakar]

@naveenjafer

Problem was my server's internet speed (maybe not speed but stability). I downloaded files with a good connection using another device. And then I transferred them to the server via external hardware.

I suggest you check your ethernet or wifi. Or of course, you can do what I did.

Basically, the code tries to extract/unzip corrupted files.

[adeljalalyousif]

/// same error

import torch
from torch import nn
import torchvision
resnet = torchvision.models.resnet101(pretrained=True)

RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory

can any one help me

[bekirbakar]

Download model manually and load it from local. See this post.

[RezaYazdaniAminabadi]

I run into this issue when different processes on different GPU ranks want to load the same file using map_location='cpu'. I have also checked that I can load the same checkpoint at python shell. Does anyone know how to resolve this?
Thanks,
Reza

[RezaYazdaniAminabadi]

By the way, it would be nice if we can get the error message shows the name of the file and the rank that cannot load the checkpoint!

[QinHsiu]

I have faced the same problem.