incompatibility of the tensorflow version with the numpy version

Question

incompatibility of the tensorflow version with the numpy version

rania-o opened this issue a year ago · comments

Hi developer,

I tried to install CHEUI and its dependencies, but when I installed tensorflow-gpu==2.4.1, numpy is downgraded to version1.19.5.
Is it normal that the versions of tensorflow-gpu and numpy that you're recommending are not compatible ?
I haven't tested CHEUI yet, I hope it won't cause any errors when running the different scripts.

Thank you,
Rania

Eduardo Eyras · Answer 1 · Sat Feb 18 2023 14:44:07 GMT+0800 (China Standard Time)

Hi,

it should be fine if the TensorFlow is installed properly.

Best

E.

rania-o · Answer 2 · Tue Feb 21 2023 23:19:03 GMT+0800 (China Standard Time)

Hi @EduEyras

Thanks for your fast reply.
I have another question, I would like to test CHEUI and other tools on the IVT sequences you used in the paper.
I was wondering where you got the modified fast5 m6A from Jenjaroenpun, P. et al. because I did not find a m6A fast5 file in this link :
https://trace.ncbi.nlm.nih.gov/Traces/index.html?view=study&acc=SRP166020

Rania

Akanksha2511 · Answer 3 · Wed Feb 22 2023 07:04:52 GMT+0800 (China Standard Time)

Hi Rania,

Thanks for testing CHEUI. Yes, if the fast5 are not available to download from the link you will have to contact the authors.

Thanks,
Akanksha

rania-o · Answer 4 · Wed Feb 22 2023 17:46:05 GMT+0800 (China Standard Time)

Hi @Akanksha2511

Thank you for your reply.
I already did but they're not reactive :/

Rania

rania-o · Answer 5 · Fri Feb 24 2023 21:42:30 GMT+0800 (China Standard Time)

thank you @Akanksha2511

I have one last question, I saw that the reads you used for training models 1 and 2 from Liu et al were obtained with the RNA001 kit. Wouldn't this be problematic since now all the dRNA sequences (as well as the data you used for your tests) are obtained with the RNA002 kit ?

Rania

Eduardo Eyras · Answer 6 · Sun Feb 26 2023 08:32:08 GMT+0800 (China Standard Time)

Hi Rania, That's a good question. We looked into the potential effects of the kit on the signals and we could not see differences that could be due to the kit, or if there are, these are difficult to separate from the experimental variability from the nanopore sequencing. At the rate of variability and accuracy that we have, we don't see any effects. cheers Eduardo

…

On Sat, 25 Feb 2023 at 00:42, rania-o ***@***.***> wrote: thank you @Akanksha2511 <https://github.com/Akanksha2511> I have one last question, I saw that the reads you used for training models 1 and 2 from *Liu et al* were obtained with the RNA001 kit. Wouldn't this be problematic since now all the dRNA sequences (as well as the data you used for your tests) are obtained with the RNA002 kit ? Rania — Reply to this email directly, view it on GitHub <#11 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADCZKB37V6X7JBQDZG3HR3DWZC3FDANCNFSM6AAAAAAU7IWC4Q> . You are receiving this because you were mentioned.Message ID: ***@***.***>

rania-o · Answer 7 · Wed Mar 01 2023 17:32:21 GMT+0800 (China Standard Time)

Hi @EduEyras

thank you for your answer. I was curious because I saw that for Nanopolish polyA they had indeed detected a difference between the two kits [https://github.com/jts/nanopolish/issues/545], but maybe this difference only concerns the polyA tail.

Rania

pre-mRNA · Answer 8 · Fri Mar 03 2023 12:07:39 GMT+0800 (China Standard Time)

Numpy recommendation downgraded to 1.19.5. Thanks Rania!