Srijan Upadhyay

I've implemented the requested features for Learning to Rank improvements. Here's what was added: ## **Label Difference Normalization** - New parameter: `lambdarank_label_diff_normalization` (default 0.0) - When > 0, multiplies delta_metric...

Hi @nalzok, these are excellent and insightful questions about the core logic. I know this is an old issue, but I thought I'd share my understanding for anyone who comes...

Hi @meitham, jumping in on this as it's a classic and interesting focus management problem. Following up on @krassowski's excellent pointer to the keyboard interaction model documentation, I believe the...

I've also observed these anomalies in the 2025 Qatar Sprint data. Since this appears to be an upstream source issue where 'HARD' compounds are erroneously inserted, the best immediate workaround...

A possible solution to rework the logging module: 1. **Thread-Safe Central Logger**: Replace thread-local logging with a singleton logger that is safe for multi-threaded and distributed environments. 2. **Namespace-Safe API**:...

Hi `@hcho3`, thanks a lot for the feedback on my initial suggestion. You were right, a global singleton isn't the way to go, and I appreciate you pointing me in...

I'm seeing this as well. It appears to be an issue with the upstream source data not providing `TeamId` entries for this specific session, rather than a bug in the...

Hey!! I have looked into the issue and tried to resolve it, i have made relevant changes in the code and have generated a PR. Please review it once and...

Thanks for the report, @kziovas. The plan would be to: 1. Modify the `vecs.Client.__init__` to accept an optional `engine: Engine` argument. 2. Add logic inside `__init__`: * If an `engine`...

Hey @abhishek0093, Confirming the behavior described by @trivialfis. Sample weights provided to a `DMatrix` are consistently used for both the training objective and the evaluation metric. This is the intended...