Open IZIEIRIOI opened 1 year ago
Hi, @IZIEIRIOI 👋🏻! Thanks for your interest in Roboflow Notebooks. I'll try to take a look at this issue today. I'll keep you posted.
Hi @SkalskiP even I am facing the same issue from yesterday
But the issue is in different notebook. The notebook name is YOLOv4-tiny-Darknet-Roboflow.ipynb
Hi @SkalskiP, please sir fix it this bug, before I could train the yolo4-dakrnet model, but now it can't, because I'm undergo my thesis with this model
Hi @LokeshSaipureddi, then you can fix it sir in notebook name yolov4 tiny darknet ?
Hi @LokeshSaipureddi and @IZIEIRIOI! Sorry, it takes so long.
@Jacobsolawetz, I could use your help here! 🥺
It looks like we have problems with installing the darknet on Google Colab on T4 GPU. I think Google must have updated Colab environment. But I never really worked with Darknet. Could you take a look?
@Jacobsolawetz, please sir can you fix it this bug ?
@SkalskiP Its not issue with K80 GPU. Colab has changed its gpu to Tesla T4 from Tesla K80. So, Installation was not working in Tesla T4 GPU
In this Image darknet file was installed. That file was not installed which was causing the error. That file was missing.
@LokeshSaipureddi yep your right, i try to install then this happen, error --"
@IZIEIRIOI
Line 5 is commented in this cell try uncommenting it. Then you will see error but you can see darknet file will be created in darknet folder. This will work !!!.
@SkalskiP I want you to update this notebook accordingly. Thank you !!!!
@LokeshSaipureddi did you tested the rest of the notebook? Does the training work after that change?
Yes training is working after that change.
I’m off today. I’ll test the fix tomorrow and if it will work then I release the update.
@LokeshSaipureddi I just tested that solution, and it doesn't work for me. Could you send me the version of the notebook that worked for you?
It doesn't work for me either after doing that change but I also did receive an error in the console when I was installing Darknet for YOLOv4 on Colab - "Makefile:162: recipe for target 'obj/convolutional_layer.o' failed" I have no clue what that means but I am letting you know that I get that when I run the step
@Strizerz, that fix does not work for me either. That bug is a bit problematic as YOLOv4 does not mention that project anymore. We will try to fight it regardless.
The same problem continues.
it doesn't work,
Hi, @fatihcihant and @seyedamirhh! 👋🏻 I'm really sorry it takes so long, but solving that issue is more complex. YOLOv4 is no longer supported. And there is no team on the other side to help us solve this problem.
Did you solve the Problem? Because in the Last week it doesn`t work and for some reason yesterday it worked ! And today with the same Code there is again the bin/bash Error
I got the same Problem with YoloV3. Is the Issue also known there?
@wolverineV2, unfortunately, that issue is still active. We did not manage to solve it. s for YOLOv3 - anything that uses Darknet will suffer from that issue. The biggest problem is that no one is supporting those models anymore. :/
Is the issue solved? I am facing the same issue.
Unfortunately not yet.
I suggest Roboflow can refet to below and modify your repo. This will build reputation of your product. Otherwise, many people will feel frustrated after following your tutorial video about building custom yolo v4 model.
https://colab.research.google.com/drive/1zqRb08ljHvIIMR4fgAXeNy1kUtjDU85B?usp=sharing
Search before asking
Notebook name
https://www.youtube.com/redirect?event=video_description&redir_token=QUFFLUhqbnJERWU5czNTUjJfMFZoTmM4ZHNFcnJsZzdqQXxBQ3Jtc0trSHRNdlRMVm1hLVZuMkpnMV9MR192TXVVYnRuOW9lVjZKa1pNeWFlSi1TSzFQNmJ0ZXR3WnU5SXZheUQ1dXNrQ3dEZjh3c2FBTGhTU3FlRWpRZDhPTURPUnZDRXFHVWtVamtmYWgyX2dRVE1ONUFhQQ&q=https%3A%2F%2Fcolab.research.google.com%2Fdrive%2F1mzL6WyY9BRx4xX476eQdhKDnd_eixBlG%23scrollTo%3DGNVU7eu9CQj3&v=N-GS8cmDPog
Bug
Environment
Minimal Reproducible Example
!./darknet detector train data/obj.data cfg/custom-yolov4-detector.cfg yolov4.conv.137 -dont_show -map
If you get CUDA out of memory adjust subdivisions above!
adjust max batches down for shorter training above
Additional
No response
Are you willing to submit a PR?