search

juglab / n2v

This is the implementation of Noise2Void training.
Other
387 stars 107 forks source link

history = model.train() won't work due to dtype float 32 #93

Closed citypalmtree closed 4 years ago

citypalmtree commented 4 years ago

Hello.

I am using n2v in python following the examples in this Github. And I have been able to train and predict using n2v.

However, for some reason, I cannot train this specific data set anymore. It gives me an error in:

ValueError Traceback (most recent call last)

in () 1 # We are ready to start training now. ----> 2 history = model.train(X, X_val) 18 frames /tensorflow-1.15.2/python3.6/tensorflow_core/python/framework/ops.py in internal_convert_to_tensor(value, dtype, name, as_ref, preferred_dtype, ctx, accepted_result_types) 1271 raise ValueError( 1272 "Tensor conversion requested dtype %s for Tensor with dtype %s: %r" % -> 1273 (dtype.name, value.dtype.name, value)) 1274 return value 1275 ValueError: Tensor conversion requested dtype float32_ref for Tensor with dtype float32: X and X_val have both dtype=float32 I compared images in this data set with other images that work in n2v, and noticed not a single difference. Both images are 32 bit images and have similar pixel ranges. Has anyone had this issue? Any advice would be greatly appreciated!
tibuch commented 4 years ago

Hi @citypalmtree

This is a first!

Would you be able to provide us with a minimal ‘working’ example showcasing the bug? That would help us tracking down the problem.

I will tag @turekg, maybe she has time to look into this 😇 🙏 .

citypalmtree commented 4 years ago

Hi @tibuch ,

I realized that this issue is because I wasn't connected to GPU... I was using Google Colab, and I didn't notice that my GPU access has been terminated after 12 hours of use. Sorry about this!

I am having another issue this time, and I think it's related to Tensorboard. I will open another issue.

Thank you for your response!