Tutorial Overview

This section focuses on practical Torch-RecHub usage patterns across different recommendation scenarios. All code snippets in this section assume you are using the sample data included in the repository and running from the repository root.

Code resources

• Full Python example scripts: examples/
• Step-by-step tutorials in the docs: docs/en/tutorials/

Tutorial List

Tutorial	Best for	Link
CTR Prediction	Ranking / click-through rate prediction	CTR tutorial
Retrieval Models	Two-tower retrieval / vector search	Retrieval tutorial
Multi-Task Learning	Joint CTR/CVR modeling	Multi-task tutorial

Quick Navigation

CTR Prediction (Ranking)

Best if you want to quickly run through WideDeep / DeepFM / DCN.

from torch_rechub.models.ranking import DeepFM
from torch_rechub.trainers import CTRTrainer

model = DeepFM(deep_features=deep_features, fm_features=fm_features, mlp_params={"dims": [256, 128]})
trainer = CTRTrainer(model, device="cuda:0")
trainer.fit(train_dl, val_dl)

View full tutorial →

Retrieval Models

Best if you want to run a two-tower or multi-interest retrieval pipeline such as DSSM / YoutubeDNN / MIND.

from torch_rechub.models.matching import DSSM
from torch_rechub.trainers import MatchTrainer

model = DSSM(user_features, item_features)
trainer = MatchTrainer(model, mode=0, device="cuda:0")
trainer.fit(train_dl)

View full tutorial →

Multi-Task Learning

Best if you want to understand the training flow of MMOE / PLE / ESMM on the Ali-CCP sample data.

from torch_rechub.models.multi_task import MMOE
from torch_rechub.trainers import MTLTrainer

model = MMOE(
    features,
    task_types=["classification", "classification"],
    n_expert=8,
    expert_params={"dims": [16]},
    tower_params_list=[{"dims": [8]}, {"dims": [8]}],
)
trainer = MTLTrainer(model, task_types=["classification", "classification"], device="cuda:0")
trainer.fit(train_dl, val_dl)

View full tutorial →

Model Tutorials

The model-specific tutorial pages below provide focused walkthroughs, including model setup, trainer usage, and tuning-oriented notes for each family.

Ranking Models

Model	Summary	Link
DeepFM	FM + deep network for ranking	DeepFM
Wide&Deep	Memorization + generalization	WideDeep
DCN / DCNv2	Explicit feature crossing	DCN
DIN	Target-aware attention over user history	DIN
DIEN	Interest evolution modeling	DIEN
BST	Transformer-based sequence ranking	BST

Retrieval Models

Model	Summary	Link
DSSM	Classic two-tower semantic matching	DSSM
YoutubeDNN	YouTube-style deep retrieval	YoutubeDNN
MIND	Multi-interest retrieval with capsules	MIND

Multi-Task Models

Model	Summary	Link
MMOE	Multi-gate mixture-of-experts	MMOE
PLE	Progressive layered extraction	PLE

Suggested Validation Order

1. Start with Quick Start to verify that your environment, trainers, and sample datasets are all working.
2. Then read the CTR tutorial or Retrieval tutorial to understand the full data flow.
3. Finally, go deeper into model-specific pages for parameter explanations, tuning suggestions, and ONNX / visualization usage.