|
|
--- |
|
|
license: cc-by-nc-4.0 |
|
|
language: |
|
|
- en |
|
|
library_name: transformers |
|
|
--- |
|
|
# RecGPT: Generative Pre-training for Text-based Recommendation |
|
|
|
|
|
We present the first domain-adapted and fully-trained large language model, RecGPT-7B, and its instruction-following variant, RecGPT-7B-Instruct, for text-based recommendation. Experimental results on rating prediction and sequential recommendation tasks show that our model, RecGPT-7B-Instruct, outperforms previous strong baselines. The general architecture and experimental results of RecGPT can be found in our [paper](arxivlink): |
|
|
|
|
|
``` |
|
|
@inproceedings{RecGPT, |
|
|
title = {{RecGPT: Generative Pre-training for Text-based Recommendation}}, |
|
|
author = {Hoang Ngo and Dat Quoc Nguyen}, |
|
|
booktitle = {Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics}, |
|
|
year = {2024} |
|
|
} |
|
|
``` |
|
|
|
|
|
We publicly release the RecGPT models along with their pre-training and fine-tuning datasets. |
|
|
|
|
|
Please cite our paper whenever RecGPT or the datasets are used to help produce published results or are incorporated into other software. |
|
|
|
|
|
For further information or requests, please go to [RecGPT's homepage](https://github.com/VinAIResearch/RecGPT)! |
|
|
|
|
|
## Usage |
|
|
```python |
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer |
|
|
|
|
|
model_path = "vinai/RecGPT-7B-Instruct" |
|
|
|
|
|
tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True) |
|
|
model = AutoModelForCausalLM.from_pretrained(model_path, device_map='auto') |
|
|
|
|
|
prompts = [ |
|
|
"### Instruction:\nPredict the next item.\nGiven the interaction history of a user with food recipes as follows:\n\nTitle: white cheddar potato gratin\nTitle: possum s cream cheese chocolate chip bread\nTitle: zucchini lasagna lasagne low carb\nTitle: sweet potato and apple casserole\n\n### Response:", |
|
|
"### Instruction:\nPredict rating for the last item.\nGiven the interaction history of a user with books as follows:\n\nTitle: Rebooting My Brain: How a Freak Aneurysm Reframed My Life; Author: Visit Amazon's Maria Ross Page; Review: I would recommend this book to everyone! As I read the last page I felt that I wanted it to continue with anniversary updates. There is so much hope wry yen within the pages.; Rating: 5.0/5.0\nTitle: Summer of Firefly Memories (The Loon Lake Series); Author: Visit Amazon's Joan Gable Page; Review: Not as well written or edited as I would have liked but the story did what the author hoped it would do---gave you pause to reflect your own 'firefly memories'. this makes it a success. I look forward to the sequel,; Rating: 5.0/5.0\nTitle: Jon Stewart: Beyond The Moments Of Zen; Author: Visit Amazon's Bruce Watson Page; Review: Jon Stewart is a gift to the ages. Thank you ,Thank you,Thank you. My ears are never tired! long may he rage.; Rating: 5.0/5.0\nTitle: How to Succeed at Aging Without Really Dying; Author: Visit Amazon's Lyla Blake Ward Page; Review: Wonderful, funny, accurate!; Rating: 5.0/5.0\nTitle: Death of an Expert Witness (Adam Dalgliesh Mystery Series #6); Author: Visit Amazon's P. D. James Page; Review: You can always depend on a good read from PD James!; Rating:\n\n### Response:", |
|
|
] |
|
|
|
|
|
for prompt in prompts: |
|
|
input_ids = tokenizer(prompt, add_special_tokens=True, return_tensors="pt")["input_ids"].to(model.device) |
|
|
output_ids = model.generate( |
|
|
input_ids, |
|
|
max_new_tokens=128, |
|
|
eos_token_id=tokenizer.eos_token_id, |
|
|
pad_token_id=tokenizer.pad_token_id, |
|
|
) |
|
|
response = tokenizer.batch_decode(output_ids[:, input_ids.size(1) :], skip_special_tokens=True)[0].strip() |
|
|
print(response) |
|
|
|
|
|
``` |
