zjkarina/omniRecsysLLM_idmodality

Recommendation model based on ID modality for Amazon Fashion.

Description

This model uses item ID embeddings to generate recommendations based on users’ purchase history.

Architecture

• Base model: Qwen2.5-Omni-7B
• Item vocabulary size: 709,036
• ID embedding dimension: 512
• Fusion head dimension: 1024
• Dataset: Amazon Fashion 2023 Full

Использование

from any2any_trainer.models.recommendation import RecommendationModel

# Load model
model = RecommendationModel.from_pretrained("zjkarina/omniRecsysLLM_idmodality")

# Generate recommendations
recommendations = model.predict_next_item(
    text="The user bought jeans and a t-shirt",
    id_ids=[12345, 67890],  # Item IDs from purchase history
    top_k=5
)

Training

The model was trained on the Amazon Fashion 2023 dataset using the ID modality.