TencentAdvertisingAlgorithmCompetition

TencentAdvertisingAlgorithmCompetition / baseline_2025

Public

This is the official baseline code for TAAC2025(parquet format).

44
6
100% credibility
Found Apr 10, 2026 at 44 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Open-source baseline code for the 2025 Tencent Advertising Algorithm Competition to train sequence-based models for personalized ad recommendations from user behavior data.

How It Works

1
🔍 Discover the competition starter kit

You hear about the Tencent Ads contest and find this ready-made helper to jumpstart your ad recommendation predictions.

2
📥 Grab the kit and your data

Download everything and place your user behavior and ad info into a simple folder.

3
🛠️ Set up the helpers

Run one easy command to get all the supporting tools ready to go.

4
🚀 Train your smart recommender

Hit start and watch it learn patterns from user histories to suggest the best ads.

5
🧪 Test and tweak for better results

Check how well it predicts, make small changes, and train again to improve.

6
📊 Generate your predictions

Run the final check on new data to get lists of top ad matches for each user.

🏆 Ready for the contest submission

You now have strong ad recommendations to enter the competition and compete for top scores.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 44 to 44 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is baseline_2025?

This Python repo delivers the official baseline 2025 for Tencent's Advertising Algorithm Competition, processing parquet data to train Transformer models on user behavior sequences for next-item prediction in ads recsys. It handles sparse features, multimodal embeddings, and generates user embeddings for top-10 retrieval via Faiss ANN search. Run training or inference by setting env vars like TRAIN_DATA_PATH and firing python main.py or eval.py.

Why is it gaining traction?

Stands out as the baseline official GitHub release with dead-simple pip setup and env-driven paths—no YAML hell or Docker dance. Developers dig the ready Faiss integration for production-scale retrieval and RQ-VAE hooks for semantic IDs from multimodal data. Even with just 44 stars, it's the baseline 2025 go-to for competition baselines.

Who should use this?

Competitors tackling Tencent Ads Algorithm Competition 2025, needing a quick sequential recsys baseline on parquet user histories and item feats. Recsys teams prototyping Transformer models with embeddings, especially those integrating Faiss for recall pipelines.

Verdict

Grab this official baseline 2025 if you're in the competition—clear docs in dual languages make it runnable out-of-box, despite 1.0% credibility score, low stars, and zero tests signaling early maturity. Fork and beat the baseline fast.

(178 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.