Move README.md
Browse files- benchmarks/README.md +124 -0
benchmarks/README.md
ADDED
@@ -0,0 +1,124 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Reproducibility Guide
|
2 |
+
|
3 |
+
## Overview
|
4 |
+
|
5 |
+
This part of repo contains the implementation and experiments. This guide will help you reproduce the results using Docker or manual installation.
|
6 |
+
|
7 |
+
---
|
8 |
+
|
9 |
+
## Docker Setup (Recommended)
|
10 |
+
|
11 |
+
### 1. Build Docker Image
|
12 |
+
|
13 |
+
```bash
|
14 |
+
docker build -t yambda-image .
|
15 |
+
```
|
16 |
+
|
17 |
+
### 2. Run Container with GPU Support
|
18 |
+
|
19 |
+
```bash
|
20 |
+
docker run --gpus all \
|
21 |
+
--runtime=nvidia \
|
22 |
+
-it \
|
23 |
+
-v </absolute/path/to/local/data>:/yambda/data \
|
24 |
+
yambda-image
|
25 |
+
```
|
26 |
+
|
27 |
+
---
|
28 |
+
|
29 |
+
## Data Organization
|
30 |
+
|
31 |
+
Create following structure in mounted data directory:
|
32 |
+
|
33 |
+
```bash
|
34 |
+
data/
|
35 |
+
βββ flat/
|
36 |
+
β βββ 50m/
|
37 |
+
β βββ likes.parquet
|
38 |
+
β βββ listens.parquet
|
39 |
+
β βββ ...
|
40 |
+
βββ sequential/
|
41 |
+
βββ 50m/
|
42 |
+
βββ likes.parquet
|
43 |
+
βββ listens.parquet
|
44 |
+
βββ ...
|
45 |
+
```
|
46 |
+
|
47 |
+
Note:
|
48 |
+
Sequential data is only needed for sasrec. You can build it from flat using scripts/transform2sequential.py or download
|
49 |
+
|
50 |
+
---
|
51 |
+
|
52 |
+
## Running Experiments
|
53 |
+
|
54 |
+
### General Usage
|
55 |
+
|
56 |
+
```bash
|
57 |
+
# For example random_rec
|
58 |
+
|
59 |
+
cd models/random_rec/
|
60 |
+
|
61 |
+
# Show help for main script
|
62 |
+
python main.py --help
|
63 |
+
|
64 |
+
# Basic execution
|
65 |
+
python main.py
|
66 |
+
```
|
67 |
+
|
68 |
+
### Specific Methods
|
69 |
+
|
70 |
+
#### BPR/ALS
|
71 |
+
|
72 |
+
```bash
|
73 |
+
cd models/bpr_als
|
74 |
+
|
75 |
+
python main.py --model bpr
|
76 |
+
python main.py --model als
|
77 |
+
```
|
78 |
+
|
79 |
+
#### SASRec
|
80 |
+
|
81 |
+
```bash
|
82 |
+
cd models/sasrec
|
83 |
+
|
84 |
+
# Training
|
85 |
+
python train.py --exp_name exp1
|
86 |
+
|
87 |
+
# Evaluation
|
88 |
+
python eval.py --exp_name exp1
|
89 |
+
```
|
90 |
+
---
|
91 |
+
|
92 |
+
## Manual Installation (Not Recommedned)
|
93 |
+
|
94 |
+
### 1. Install Core Dependencies
|
95 |
+
|
96 |
+
```bash
|
97 |
+
pip install torch torchvision torchaudio
|
98 |
+
```
|
99 |
+
|
100 |
+
### 2. Install Implicit (CUDA 11.8 required)
|
101 |
+
|
102 |
+
Implicit works only with cuda<12. See reasons [here](https://github.com/NVIDIA/nvidia-docker/issues/700#issuecomment-381073278)
|
103 |
+
|
104 |
+
```bash
|
105 |
+
CUDACXX=/usr/local/cuda-11.8/bin/nvcc \
|
106 |
+
pip install implicit
|
107 |
+
```
|
108 |
+
|
109 |
+
### 3. Install SANSA
|
110 |
+
|
111 |
+
```bash
|
112 |
+
sudo apt-get install libsuitesparse-dev
|
113 |
+
git clone https://github.com/glami/sansa.git
|
114 |
+
cd sansa && \
|
115 |
+
SUITESPARSE_INCLUDE_DIR=/usr/include/suitesparse \
|
116 |
+
SUITESPARSE_LIBRARY_DIR=/usr/lib \
|
117 |
+
pip install .
|
118 |
+
```
|
119 |
+
|
120 |
+
### 4. Install Project Package
|
121 |
+
|
122 |
+
```bash
|
123 |
+
pip install . # In root directory
|
124 |
+
```
|