Bug: 403 CORS error -tfhub/imagenet/mobilenet. Issue was last night, resolved this morning. Occurred again 3/26 1:47pm est

[devihall03]

What happened?

Around 9pm yesterday had an issue with imagenet/mobilenet_v1_025_224/classification/1/model.json

Webapp made api call to
https://storage.googleapis.com/tfjs-models/tfjs/mobilenet_v1_0.25_224/model.json

got redirected to
https://storage.googleapis.com/kagglesdsdata/models/1514/1807/model.json?X-Goog-Algorithm=GOOG4-RSA-SHA256&X-Goog-Credential=gcp-kaggle-com%40kaggle-161607.iam.gserviceaccount.com%2F20240322%2Fauto%2Fstorage%2Fgoog4_request&X-Goog-Date=20240322T025731Z&X-Goog-Expires=259200&X-Goog-SignedHeaders=host&X-Goog-Signature=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

Model did not load and console showed
error: Access to fetch at 'https://tfhub.dev/google/imagenet/mobilenet_v1_025_224/classification/1/model.json?tfjs-format=file' from origin 'http://127.0.0.1:5501' has been blocked by CORS policy: The 'Access-Control-Allow-Origin' header has a value 'http://127.0.0.1:5500' that is not equal to the supplied origin. Have the server send the header with a valid value, or, if an opaque response serves your needs, set the request's mode to 'no-cors' to fetch the resource with CORS disabled.

IP address had not changed. Using the same device. One minute requesta and response was fine, next minute CORS error.

Relevant code

// Function to handle model loading
function modelReady() {
  console.log("Model is ready");
}

Relevant log output

error: Access to fetch at 'https://tfhub.dev/google/imagenet/mobilenet_v1_025_224/classification/1/model.json?tfjs-format=file' from origin 'http://127.0.0.1:5501' has been blocked by CORS policy: The 'Access-Control-Allow-Origin' header has a value 'http://127.0.0.1:5500' that is not equal to the supplied origin. Have the server send the header with a valid value, or, if an opaque response serves your needs, set the request's mode to 'no-cors' to fetch the resource with CORS disabled.

tensorflow_hub Version

0.12.0 (latest stable release)

TensorFlow Version

2.8 (latest stable release)

Other libraries

ML5
P5
JQuery
Bootstrap

Python Version

3.x

OS

macOS

[KeijiBranshi]

Hi @devihall03, thanks for posting.

Do you mind reproducing and sharing the HTTP Request-Response headers associated with your failed requests to https://tfhub.dev/google/imagenet/mobilenet_v1_025_224/classification/1/model.json?tfjs-format=file?

Here are the instructions to find them on Chrome, but it may differ depending on your browser.

Thanks!

[devihall03]

General:
Request URL:
https://storage.googleapis.com/tfjs-models/tfjs/mobilenet_v1_0.25_224/group43-shard1of1
Request Method:
GET
Status Code:
200 OK (from disk cache)
Remote Address:
[2607:f8b0:4002:c05::cf]:443
Referrer Policy:
strict-origin-when-cross-origin

Response Headers:
Accept-Ranges:
bytes
Access-Control-Allow-Origin:
*
Access-Control-Expose-Headers:
*, Content-Length, Date, Server, Transfer-Encoding, X-GUploader-UploadID, X-Google-Trace
Age:
0
Alt-Svc:
h3=":443"; ma=2592000,h3-29=":443"; ma=2592000
Cache-Control:
public, max-age=3600
Content-Length:
512
Content-Type:
application/octet-stream
Date:
Tue, 26 Mar 2024 17:29:40 GMT
Etag:
"f98a7ecec81976b48509940e9a0ce366"
Expires:
Tue, 26 Mar 2024 18:29:40 GMT
Last-Modified:
Mon, 12 Mar 2018 18:28:21 GMT
Server:
UploadServer
X-Goog-Generation:
1520879301195712
X-Goog-Hash:
crc32c=NPQ3bw==, md5=+Yp+zsgZdrSFCZQOmgzjZg==
X-Goog-Metageneration:
2
X-Goog-Storage-Class:
MULTI_REGIONAL
X-Goog-Stored-Content-Encoding:
identity
X-Goog-Stored-Content-Length:
512
X-Guploader-Uploadid:
ABPtcPqW00WYCwiRgkjOmxeYVQnGDSuvEMLG3OevNGgFNhZHJHKkmB8bPyNH8050NcnYMDmNgFg3EFsL2A

Request Headers:
Referer:
http://127.0.0.1:5500/
Sec-Ch-Ua:
"Google Chrome";v="123", "Not:A-Brand";v="8", "Chromium";v="123"
Sec-Ch-Ua-Mobile:
?0
Sec-Ch-Ua-Platform:
"macOS"
User-Agent:
Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/123.0.0.0 Safari/537.36

[devihall03]

[devihall03]

@KeijiBranshi Just got this same error again just now. What is going on??? It was working one minute and now its not!

[KeijiBranshi]

Hi @devihall03. Looked into this a little more.

The logs output from your original post shows your client using two separate ports http://127.0.0.1:5501 and http://127.0.0.1:5500:

error: Access to fetch at 'https://tfhub.dev/google/imagenet/mobilenet_v1_025_224/classification/1/model.json?tfjs-format=file' from origin 'http://127.0.0.1:5501' has been blocked by CORS policy: The 'Access-Control-Allow-Origin' header has a value 'http://127.0.0.1:5500' that is not equal to the supplied origin. Have the server send the header with a valid value, or, if an opaque response serves your needs, set the request's mode to 'no-cors' to fetch the resource with CORS disabled.

Kaggle's servers previously just mirrored the requestor's origin in the response headers. I don't understand how, but your tfhub.dev requests might have been sent on one port and handled on another.

I just adjusted the CORS response header to be a * for these tfjs requests. I'm hoping that mitigates this issue, but please post back if you encounter the CORS error again.

[devihall03]

@KeijiBranshi Was working great and it happened again! Just before an important demo!

Is there anything that can be done to mitigate this???

[devihall03]

http://localhost:4000/ is our local

[KeijiBranshi]

Hi, we're looking into this issue. Thanks for your patience.

[KeijiBranshi]

Hi. We rolled out a fix. But please keep up posted if you see this again.

More info here: #908 (comment)

Thanks!

[KeijiBranshi]

Apologies that this happened before an important demo. For situations like this, I'd highly recommend self-hosting the model on your site. You can download the model directly from Kaggle:

• Model Page: https://www.kaggle.com/models/google/mobilenet-v1/tfJs/025-224-classification
• Direct Download: https://www.kaggle.com/api/v1/models/google/mobilenet-v1/tfJs/025-224-classification/2/download

Alternatively, you can bundle TF.js models directly into your web apps:

• https://www.tensorflow.org/js/tutorials/deployment/size_optimized_bundles

For more support on that, I'd recommend posting in the tensorflow repository since this repo is specifically for the Python tfhub client.

[KeijiBranshi]

Thanks again for sharing. Since the issue is resolved, I'm going to close these related threads out (#908, #909, #904).

[google-ml-butler]

Are you satisfied with the resolution of your issue?
Yes
No