cocoshe

cocoshe / MIMIGenRec

Public

A Flexible Framework for Generative Recommendation

24
7
100% credibility
Found Feb 22, 2026 at 17 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

MIMIGenRec is a framework for training AI models that generate personalized product recommendations using techniques like supervised fine-tuning and reinforcement learning on datasets such as Amazon reviews.

How It Works

1
🔍 Discover MIMIGenRec

You stumble upon this tool while searching for ways to build smart recommendation systems that suggest products like Amazon does.

2
📦 Get started quickly

Download and set it up with a simple command, so your computer is ready in minutes.

3
📊 Grab sample shopping data

Pull real-world purchase histories from public sources to experiment with right away.

4
🔧 Prepare your data

Transform the data into a format the tool understands, creating unique codes for items.

5
🚀 Train your first recommender

Run the basic training to teach it patterns from past buys.

6
⚡ Boost with smart feedback

Add advanced learning that rewards good suggestions, making recommendations sharper and more accurate.

7
📈 Test and measure success

Generate suggestions and check scores like hit rates to see how well it performs.

🎉 Your recommender shines

Enjoy personalized product lists that delight users, ready for real apps.

Sign up to see the full architecture

6 more

Sign Up Free

Star Growth

See how this repo grew from 17 to 24 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 MIMIGenRec?

MIMIGenRec is a Python framework for training generative recommendation models that predict next items as sequences of semantic IDs, constrained to valid outputs via trie-based decoding during beam search. It handles supervised fine-tuning on interaction histories and reinforcement learning with ranking rewards like NDCG, starting from Amazon datasets or custom data. Users get CLI scripts for data prep, SFT via YAML configs, RL policy optimization, and evaluation metrics like HR@K and NDCG@K.

Why is it gaining traction?

It stands out by integrating Hugging Face tools like LlamaFactory for easy LoRA/SFT on small models (0.5B-3B) and TRL for multi-GPU RL with DeepSpeed ZeRO, without needing complex recsys libraries. Developers appreciate the quick-start pipeline: download data, preprocess to JSON, train, and evaluate in hours, plus flexibility for custom rewards and datasets. The constrained generation ensures realistic recommendations, hooking recsys folks tired of embedding-based methods.

Who should use this?

Recsys engineers building sequential recommenders on e-commerce or content data, especially those with item metadata for semantic IDs. Researchers prototyping LLM-based next-item prediction or RLHF for ranking. Teams evaluating generative recsys on benchmarks like Amazon Industrial_and_Scientific.

Verdict

Grab it for flexible generative recsys experiments if you have GPU clusters—solid HF integration makes prototyping fast. With 18 stars and 1.0% credibility, it's early-stage; docs cover basics well but expect tweaks for production.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.