the filepath provided must end in `.weights.h5`

[schienstockd]

Hello,

I'm using tensorflow 2.16.1 and get the following when trying to train a model with model.train

ValueError: When using `save_weights_only=True` in `ModelCheckpoint`, the filepath provided must end in `.weights.h5` (Keras weights format). Received: filepath=*/model_name/weights_best.h5

I had no issue with this before. Is this a new checkpoint in keras?

Thanks!

[jdeschamps]

Hi,

Sorry for the delay!

Yes it seems to be a problem of the newer TF... Can you downgrade to an earlier version?

We tested it last with TF2.13 and TF2.10. On CPU just now, it seems it works with TF2.15.

[jdeschamps]

I have updated the installation instructions.

[XarlesMicro]

should not the instructions read:
python3 -m pip install tensorflow==2.10
Otherwise, it will default to the latest version, which will not work. I just had the same problem until:
pip uninstall tensorflow
python3 -m pip install tensorflow==2.10

[romydhan]

Hi, I'm using TF2.10 and still getting this error....is there a fix?

[jdeschamps]

That's surprising, I don't really remember but if I trust my past-self, this specific error should not appear with TF 2.10. Could you paste the exact error you are getting?

Please note that we do not intend to continue maintaining n2v and we will retire it soon. Instead, we are focusing on CAREamics, which allows you to run Noise2Void as well. If you encounter issues with CAREamics, we are more likely to fix them right away. 😃

That being said, I will try to issue a fix for this current issue, as it is simple enough to patch.

[romydhan]

Thanks for the quick response. FYI, I followed the instructions posted here to install everything but it's possible I did something wrong. This is the error I'm getting:
ValueError: When using save_weights_only=True in ModelCheckpoint, the filepath provided must end in .weights.h5 (Keras weights format). Received: filepath=C:\Users\CT-romy.napari\N2V\models\n2v_2D\weights_best.h5

Earlier on I see this, if this is helpful:
INFO: Creating model
10:32:36 : WARNING : Dummy-1 : From C:\ProgramData\anaconda3\envs\napari-n2v\lib\site-packages\keras\src\backend\tensorflow\core.py:222: The name tf.placeholder is deprecated. Please use tf.compat.v1.placeholder instead.

2024-12-05 10:32:36.595188: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX512F AVX512_VNNI AVX512_BF16 AVX512_FP16 AVX_VNNI AMX_TILE AMX_INT8 AMX_BF16 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.

I'm actually trying to just verify the TensorFlow version and having issues doing that. I'm new to it.

Thanks.

[jdeschamps]

To be fair to you, the instructions are a bit of a mess!

Can you type in your console:

conda list | grep tensor

That should print the tensorflow version in your conda environment!

[romydhan]

I'm on windows, but when I type this

"conda list | findstr tensor"

it returns nothing....

[looking into CAREamics now, might be quicker to get that working than solve this issue]

[jdeschamps]

Ah sorry! We usually advise people to use the "Git for windows" terminal under Windows, because then they can use Unix commands.

Then maybe just conda list and check out manually tensorflow in the list. You need to make sure that your conda environment is active.

[looking into CAREamics now, might be quicker to get that working than solve this issue]

Great, don't hesitate to open an issue there then!

[VANC-five]

I got the same error...
The filename must end in .weights.h5. Received: filepath=./models/simulation_study1/n_persons_100/n_items_30/rep_0/net_depth_1/linear_True\person_net/person_net

[jdeschamps]

Hi,

Which TensorFlow version are you using?

We are not maintaining n2v anymore, I would suggest using CAREamics instead to run Noise2Void, N2V2, or structN2V.