heycao commited on
Commit
82610df
·
verified ·
1 Parent(s): 6474213

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +169 -0
README.md ADDED
@@ -0,0 +1,169 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - recommender-system
4
+ - sasrec
5
+ datasets:
6
+ - yelp
7
+ ---
8
+
9
+ # SASRec - yelp
10
+
11
+ ## Model Description
12
+ SASRec (Self-Attentive Sequential Recommendation) is a sequential recommendation model that uses self-attention mechanisms to model user's historical behavior data.
13
+
14
+ ## Training Results
15
+ ### Best Valid Results
16
+ - **ndcg@10**: 0.033
17
+ - **hit@10**: 0.0638
18
+ - **averagepopularity@10**: 303.7884
19
+ - **giniindex@10**: 0.8711
20
+ - **itemcoverage@10**: 0.7292
21
+ - **shannonentropy@10**: 0.0002
22
+ - **tailpercentage@10**: 0.0087
23
+
24
+ ### Test Results
25
+ - **ndcg@10**: 0.0325
26
+ - **hit@10**: 0.0613
27
+ - **averagepopularity@10**: 296.3756
28
+ - **giniindex@10**: 0.8748
29
+ - **itemcoverage@10**: 0.7134
30
+ - **shannonentropy@10**: 0.0002
31
+ - **tailpercentage@10**: 0.0105
32
+
33
+ ## Configuration
34
+ The model was trained with the following configuration:
35
+ ```yaml
36
+ ENTITY_ID_FIELD: entity_id
37
+ HEAD_ENTITY_ID_FIELD: head_id
38
+ ITEM_ID_FIELD: item_id
39
+ ITEM_LIST_LENGTH_FIELD: item_length
40
+ LABEL_FIELD: label
41
+ LIST_SUFFIX: _list
42
+ MAX_ITEM_LIST_LENGTH: 50
43
+ MODEL_INPUT_TYPE: InputType.POINTWISE
44
+ MODEL_TYPE: ModelType.SEQUENTIAL
45
+ NEG_PREFIX: neg_
46
+ POSITION_FIELD: position_id
47
+ RATING_FIELD: rating
48
+ RELATION_ID_FIELD: relation_id
49
+ TAIL_ENTITY_ID_FIELD: tail_id
50
+ TIME_FIELD: timestamp
51
+ USER_ID_FIELD: user_id
52
+ additional_feat_suffix: null
53
+ alias_of_entity_id: null
54
+ alias_of_item_id: null
55
+ alias_of_relation_id: null
56
+ alias_of_user_id: null
57
+ attn_dropout_prob: 0.5
58
+ benchmark_filename: null
59
+ checkpoint_dir: saved
60
+ clip_grad_norm: null
61
+ dataloaders_save_path: null
62
+ dataset: yelp
63
+ dataset_save_path: null
64
+ device: cuda
65
+ discretization: null
66
+ embedding_size: 64
67
+ enable_amp: false
68
+ enable_scaler: false
69
+ entity_kg_num_interval: '[0,inf)'
70
+ epochs: 500
71
+ eval_args:
72
+ group_by: user
73
+ mode:
74
+ test: full
75
+ valid: full
76
+ order: TO
77
+ split:
78
+ LS: valid_and_test
79
+ eval_batch_size: 8192
80
+ eval_step: 1
81
+ eval_type: EvaluatorType.RANKING
82
+ field_separator: "\t"
83
+ filter_inter_by_user_or_item: true
84
+ gpu_id: '0'
85
+ hidden_act: gelu
86
+ hidden_dropout_prob: 0.5
87
+ hidden_size: 64
88
+ initializer_range: 0.02
89
+ inner_size: 256
90
+ item_inter_num_interval: '[10,inf)'
91
+ kg_reverse_r: false
92
+ layer_norm_eps: 1.0e-12
93
+ learner: adam
94
+ learning_rate: 0.001
95
+ load_col:
96
+ inter:
97
+ - user_id
98
+ - item_id
99
+ - rating
100
+ - timestamp
101
+ local_rank: 0
102
+ log_wandb: false
103
+ loss_decimal_place: 4
104
+ loss_type: CE
105
+ metric_decimal_place: 4
106
+ metrics:
107
+ - NDCG
108
+ - Hit
109
+ - AveragePopularity
110
+ - GiniIndex
111
+ - ItemCoverage
112
+ - ShannonEntropy
113
+ - TailPercentage
114
+ model: SASRec
115
+ n_heads: 2
116
+ n_layers: 2
117
+ normalize_all: null
118
+ normalize_field: null
119
+ numerical_features: []
120
+ preload_weight: null
121
+ relation_kg_num_interval: '[0,inf)'
122
+ repeatable: true
123
+ reproducibility: true
124
+ require_pow: false
125
+ rm_dup_inter: null
126
+ save_dataloaders: false
127
+ save_dataset: false
128
+ seed: 2020
129
+ seq_len: null
130
+ seq_separator: ' '
131
+ show_progress: false
132
+ shuffle: true
133
+ single_spec: true
134
+ state: INFO
135
+ stopping_step: 10
136
+ test_neg_sample_args:
137
+ distribution: uniform
138
+ sample_num: none
139
+ threshold: null
140
+ topk:
141
+ - 10
142
+ train_batch_size: 8192
143
+ train_neg_sample_args:
144
+ alpha: none
145
+ candidate_num: 0
146
+ distribution: none
147
+ dynamic: false
148
+ sample_num: none
149
+ transform: null
150
+ unload_col: null
151
+ unused_col:
152
+ inter:
153
+ - rating
154
+ use_gpu: true
155
+ user_inter_num_interval: '[10,inf)'
156
+ val_interval:
157
+ rating: '[3,inf)'
158
+ valid_metric: NDCG@10
159
+ valid_metric_bigger: true
160
+ valid_neg_sample_args:
161
+ distribution: uniform
162
+ sample_num: none
163
+ wandb_project: recbole
164
+ weight_decay: 0.0
165
+ worker: 0
166
+
167
+ ```
168
+
169
+ ## Usage