File size: 1,648 Bytes
4fc5b15 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 | ---
license: other
language:
- en
tags:
- recommendation
- information-retrieval
- feature-extraction
base_model: Qwen/Qwen3-Embedding-0.6B
---
# Amazon recommendation tools
Frozen catalog artifacts and behavior-trained recommendation tools used by RecoAtlas. This is an
inference-artifact repository, not a standalone Transformers model and not a record of training
runs.
## Categories
- `Musical_Instruments`
- `Electronics`
- `Video_Games`
## Uniform layout
Each category contains exactly the files needed by the released tools:
```text
<category>/
data/product_texts.json
data/subcategories.json
embeddings/item_embeddings_qwen.pt
embeddings/complementarity_embeddings_qwen.pt
models/retrieval_model_qwen.pt
manifest.json
```
- `retrieval_model_qwen.pt` is the selected query-side LoRA retrieval checkpoint.
- `item_embeddings_qwen.pt` contains the frozen catalog item vectors used by semantic search and
near-duplicate pruning.
- `complementarity_embeddings_qwen.pt` contains the selected catalog-aligned `anchor_embs` and
`complement_embs` used by the utility-aligned complementarity tool. It is the complementarity
inference artifact; historical epoch checkpoints and training runs are intentionally omitted.
- `manifest.json` records exact sizes and SHA-256 checksums.
The underlying query/reference benchmark is [iaouali/amazon-benchmark](https://huggingface.co/datasets/iaouali/amazon-benchmark).
The base query encoder is [Qwen/Qwen3-Embedding-0.6B](https://huggingface.co/Qwen/Qwen3-Embedding-0.6B). Catalog content,
source datasets, and the base model remain subject to their respective terms. See `NOTICE`.
|