But I guess telling the model middle name initials are not needed would be a better option. I don't know

That's fine. Thank you for sharing! I think you posted too early though. Now the board is going to be funny😂😂

😂😂 That's the goal. Let's see who can maybe be more creative than the other. Been a learning experience for me. That's a plus for me.

😂😂😭😭

😂😂 go all the way and publish the segmentation part to see real chaos

You can help us with that Nymfree🤲😂

🤣🤣🤣😂 wild

Honestly, we were also quite surprised and didn't really understand the motive to why he did that. But what is done is done. I am extremely sorry for how you feel now but it is not a wasted effort, @Muhamed_Tuo. Wishing you the best of Luck!

Well said @Muhamed_Tuo. Also found it strange as he had been sitting on this solution for around a month.

Totally get the frustration @Muhamed_Tuo.

Quick reality check on Zindi rules: "Code must not be shared privately… Any code that is shared must be made available to all participants on the discussion boards." This notebook was shared publicly, so it's within the rules.

On timing/fairness: early sharing can sting, but it raises the floor, not the ceiling. Copying a score ≠ winning-CV choices, feature tweaks, training stability, ensembling, and error analysis still separate leaders.

Upside: faster learning, more creativity, and tougher competition. A strong public baseline helps everyone optimize further-better splits/regularization/augmentation/post-processing, etc.

If we want an embargo on sharing code in the last X weeks, let's propose it for future comps. Until then, this share is rule-compliant, let's build on it, not shut it down.

@Muhamed_Tuo, massive respect for the time and effort you put in—those scores are proof of your skill, and no shared notebook can erase the learning from that journey.

For me, this is what community is all about. A shared high-performing notebook doesn't devalue the work done; it accelerates our collective growth. It shifts the challenge from baseline discovery to true innovation.

I also worked hard from the start, and I believe this is how we push the boundaries. We learn best when we build upon shared ideas.

I believe what he shared gives a baseline score which is similar to the starter notebook Zindi gives sometimes. I don't see anything wrong with it.

@Joseph_gitau The issue isn’t with sharing itself, it’s the timing. Sharing resources is great, but not this close to the end of the competition. If it had been shared a month earlier, that would’ve been fair. But right now, all the effort and resources the top competitors invested to reach those scores have basically gone to waste.

@Bone Let’s be honest the starter notebook’s score and his notebook’s score aren’t even close. Anyway, what’s done is done… let the chaos begin

May the best win!

I have got a second OCR pipeline which i will share later on. Should be better that what I have shared already but needs further checks to confirm the structure and functionality.

@Joseph_gitau Have you realized there's a bug in unsloth?

When you try any inferencing. Check on your side if I am wrong

Yes, and why I have got a second pipeline. I knew this would happen. I guess this affects many in the leaderboard. with 5 days to the close there need to be a solution for this.

Alright that's fine. I think everyone should be aware

How can we make them aware? Doing a post on the same or this thread should be ok?

Let me post on a new thread. Or?

Should be ok.

Done!