Lots of bugs in this repo

[zhangshifen38]

I have tried to run this repo in my own device with an NVIDIA GeForce RTX 3090 and python 3.11.13 with torch 2.5.1+cu121. Currently I have faced problems include:

1. Circular import: genrec/pipeline.py import genrec.utils -> genrec/utils.py import genrec.dataset -> genrec/dataset.py import genrec.utils.log

2. In config variable all string values are contained in a list, e.g. {"dataset": ["AmazonReviews2014"], ... }

3. In genrec/datasets/AmazonReviews2014/dataset.py the class AmazonReviews2014 it has no method named self._parse_gz()

4. In genrec/models/ActionPiece/ it leaks trainer so it cannot run training process, so I tried to make a new file trainer.py here and wrote 'from genrec.trainer import Trainer as ActionPieceTainer' and it worked.

5. In genrec/models/ActionPiece/model.py with the default config n_inference_ensemble=5 the assertion in line 105 failed, so in order to run I change n_inference_ensemble=4.

6. After I fixed the bugs above by myself, I run it and got this result in Sports dataset:

INFO:root:Test Results: OrderedDict([('ndcg@5', 0.01349499449133873), ('ndcg@10', 0.01731870323419571), ('ndcg@20', 0.022189823910593987), ('ndcg@50', 0.029919946566224098), ('recall@5', 0.020936375483870506), ('recall@10', 0.032879944890737534), ('recall@20', 0.05232688784599304), ('recall@50', 0.09164231270551682), ('err@5', 0.0), ('err@10', 0.0), ('err@20', 0.0), ('err@50', 0.0)])

It seems to be some problems in this repo, can anyone help me?

[hyp1231]

Hi, thanks for catching these bugs! I've fixed them and submitted a PR: #6. It should address the issues you raised, though it might take a little time to be merged.

I really appreciate the detailed report, which was super helpful!