---
tags:
- recommender
- multimodal
- amazon
- beauty
- json
- npy
- parquet
- faiss
- lora
- huggingface-dataset
---

# 📦 Amazon Beauty Subset for MMR-Agentic-CoVE

This dataset contains preprocessed files for the "Beauty" category from the Amazon Reviews dataset.
It supports the MMR-Agentic-CoVE recommender system, including FAISS indexes, LoRA-tuned model weights, and multimodal features.

Use this dataset with the backend [`cove-api`](https://huggingface.co/spaces/mickey1976/cove-api) and frontend [`cove-ui`](https://huggingface.co/spaces/mickey1976/cove-ui) for live testing.


-
license: cc-by-nc-4.0
---
Here is a revised and complete version of your README.md for the Hugging Face dataset repo mayankc-amazon_beauty_subset, reflecting the reorganized folder structure, usage examples, and links to your API/UI Spaces:

⸻


# Amazon Beauty Subset – Structured Dataset for MMR-Agentic-CoVE Recommender

This is a clean, categorized subset of the **Amazon Beauty Products Dataset** curated for the [MMR-Agentic-CoVE](https://huggingface.co/spaces/mickey1976/cove-ui) recommender system. It includes multimodal item data (text, image, metadata), user interactions, FAISS indexes, model outputs, and embedding vectors — all organized for efficient retrieval by the API and UI spaces.

## 🗂️ Folder Structure

.
├── json/        # Configs, maps, user/item sequences
├── npy/         # Embedding arrays (text, image, meta, CoVE)
├── parquet/     # Tabular structured data
├── model/       # PEFT/LoRA model weights
├── faiss/       # FAISS index files for nearest neighbor search
└── README.md

## 📁 Key Files

### `json/`
- `defaults.json`: Weight config for fusion modes
- `item_ids.json`, `user_seq.json`, `cove_item_ids.json`: ID mappings and test sets

### `npy/`
- `text.npy`, `image.npy`, `meta.npy`: Item modality embeddings
- `cove_logits.npy`, `full_cove_embeddings.npy`: CoVE model outputs

### `parquet/`
- `reviews.parquet`, `items_catalog.parquet`: Base product metadata
- `user_text_emb.parquet`: User text embedding vectors

### `model/`
- `model.safetensors`, `adapter_model.safetensors`: LoRA fine-tuned weights

### `faiss/`
- `items_beauty_concat.faiss`, `items_beauty_weighted.faiss`: FAISS indexes for fast item retrieval

---

## 🔌 Paired Spaces

- **API Backend (FastAPI):** [CoVE API](https://huggingface.co/spaces/mickey1976/cove-api)
- **UI Frontend (Gradio):** [CoVE UI](https://huggingface.co/spaces/mickey1976/cove-ui)

These Spaces dynamically fetch data from this dataset repo using `huggingface_hub`.

---

## 🐍 Example: Load Embeddings via `huggingface_hub`

```python
from huggingface_hub import hf_hub_download
import numpy as np

# Load text embeddings
text_emb_path = hf_hub_download(
    repo_id="mickey1976/mayankc-amazon_beauty_subset",
    repo_type="dataset",
    filename="npy/text.npy"
)

text_embeddings = np.load(text_emb_path)


⸻

📖 Citation

Data originally from:

Ni, J., et al. (2019). Amazon Review Dataset. UCSD.
https://nijianmo.github.io/amazon/index.html

Used here in support of MMR-Agentic-CoVE multimodal recommender architecture.

⸻

🛠 Maintained by

Mayank Choudhary
GitHub | Hugging Face

---

### ✅ Instructions to Save

1. Overwrite the current `README.md` in your dataset root directory:
```bash
nano README.md

(Paste the content above, save with Ctrl + O, exit with Ctrl + X)
	2.	Commit and push:

git add README.md
git commit -m "Update README with folder structure and usage guide"
git push


Here is a shorter version of the README.md suitable for the Hugging Face dataset card view (top-level summary users see when browsing your dataset):

⸻


# 📦 Amazon Beauty Subset for MMR-Agentic-CoVE

This dataset powers the **MMR-Agentic-CoVE** recommender system and contains a compact, multimodal slice of the Amazon Beauty product data. It includes:

- ✅ JSON configs & sequences
- ✅ NPY embeddings (text, image, meta, CoVE)
- ✅ Parquet structured tables
- ✅ PEFT model weights (LoRA/adapter)
- ✅ FAISS indexes for fast retrieval

## 🧭 Folder Structure

json/      → ID maps, defaults, sequences
npy/       → Embeddings & logits
parquet/   → Metadata & user-item tables
model/     → Fine-tuned model weights
faiss/     → Item FAISS indexes

## 🔌 Paired Spaces

- **API Backend** → [CoVE API](https://huggingface.co/spaces/mickey1976/cove-api)
- **Gradio UI** → [CoVE UI](https://huggingface.co/spaces/mickey1976/cove-ui)

## 📚 Citation

> Ni, J., et al. (2019). *Amazon Review Dataset*. UCSD.  
> https://nijianmo.github.io/amazon/index.html

Maintained by [@mickey1976](https://huggingface.co/mickey1976)


⸻