heycao commited on
Commit
0cd4f1f
·
verified ·
1 Parent(s): 7bb1a58

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +177 -0
README.md ADDED
@@ -0,0 +1,177 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - recommender-system
4
+ - sasrec
5
+ datasets:
6
+ - ML-100k
7
+ metrics:
8
+ - ndcg
9
+ - hit_rate
10
+ - averagepopularity
11
+ - giniindex
12
+ - itemcoverage
13
+ - shannonentropy
14
+ - tailpercentage
15
+ library_name: viberec
16
+ ---
17
+
18
+ # SASRec - ML-100k
19
+
20
+ ## Model Description
21
+ SASRec (Self-Attentive Sequential Recommendation) is a sequential recommendation model that uses self-attention mechanisms to model user's historical behavior data.
22
+
23
+ ## Training Results
24
+ ### Best Valid Results
25
+ - **ndcg@10**: 0.0197
26
+ - **hit@10**: 0.0362
27
+ - **averagepopularity@10**: 304.4937
28
+ - **giniindex@10**: 0.9858
29
+ - **itemcoverage@10**: 0.0747
30
+ - **shannonentropy@10**: 0.0387
31
+ - **tailpercentage@10**: 0.0002
32
+
33
+ ### Test Results
34
+ - **ndcg@10**: 0.0157
35
+ - **hit@10**: 0.0351
36
+ - **averagepopularity@10**: 304.5865
37
+ - **giniindex@10**: 0.9861
38
+ - **itemcoverage@10**: 0.0688
39
+ - **shannonentropy@10**: 0.0417
40
+ - **tailpercentage@10**: 0.0
41
+
42
+ ## Configuration
43
+ The model was trained with the following configuration:
44
+ ```yaml
45
+ ENTITY_ID_FIELD: entity_id
46
+ HEAD_ENTITY_ID_FIELD: head_id
47
+ ITEM_ID_FIELD: item_id
48
+ ITEM_LIST_LENGTH_FIELD: item_length
49
+ LABEL_FIELD: label
50
+ LIST_SUFFIX: _list
51
+ MAX_ITEM_LIST_LENGTH: 50
52
+ MODEL_INPUT_TYPE: InputType.POINTWISE
53
+ MODEL_TYPE: ModelType.SEQUENTIAL
54
+ NEG_PREFIX: neg_
55
+ POSITION_FIELD: position_id
56
+ RATING_FIELD: rating
57
+ RELATION_ID_FIELD: relation_id
58
+ TAIL_ENTITY_ID_FIELD: tail_id
59
+ TIME_FIELD: timestamp
60
+ USER_ID_FIELD: user_id
61
+ additional_feat_suffix: null
62
+ alias_of_entity_id: null
63
+ alias_of_item_id: null
64
+ alias_of_relation_id: null
65
+ alias_of_user_id: null
66
+ attn_dropout_prob: 0.5
67
+ benchmark_filename: null
68
+ checkpoint_dir: saved
69
+ clip_grad_norm: null
70
+ dataloaders_save_path: null
71
+ dataset: ml-100k
72
+ dataset_save_path: null
73
+ device: cpu
74
+ discretization: null
75
+ enable_amp: false
76
+ enable_scaler: false
77
+ entity_kg_num_interval: null
78
+ epochs: 1
79
+ eval_args:
80
+ group_by: user
81
+ mode:
82
+ test: full
83
+ valid: full
84
+ order: TO
85
+ split:
86
+ LS: valid_and_test
87
+ eval_batch_size: 40960
88
+ eval_step: 1
89
+ eval_type: EvaluatorType.RANKING
90
+ field_separator: "\t"
91
+ filter_inter_by_user_or_item: true
92
+ gpu_id: '0'
93
+ hidden_act: gelu
94
+ hidden_dropout_prob: 0.5
95
+ hidden_size: 64
96
+ initializer_range: 0.02
97
+ inner_size: 256
98
+ item_inter_num_interval: '[10,inf)'
99
+ kg_reverse_r: false
100
+ layer_norm_eps: 1.0e-12
101
+ learner: adam
102
+ learning_rate: 0.001
103
+ load_col:
104
+ inter:
105
+ - user_id
106
+ - item_id
107
+ - rating
108
+ - timestamp
109
+ local_rank: 0
110
+ log_wandb: false
111
+ loss_decimal_place: 4
112
+ loss_type: CE
113
+ metric_decimal_place: 4
114
+ metrics:
115
+ - NDCG
116
+ - Hit
117
+ - AveragePopularity
118
+ - GiniIndex
119
+ - ItemCoverage
120
+ - ShannonEntropy
121
+ - TailPercentage
122
+ model: SASRec
123
+ n_heads: 2
124
+ n_layers: 2
125
+ normalize_all: true
126
+ normalize_field: null
127
+ numerical_features: []
128
+ preload_weight: null
129
+ relation_kg_num_interval: null
130
+ repeatable: true
131
+ reproducibility: true
132
+ require_pow: false
133
+ rm_dup_inter: null
134
+ save_dataloaders: false
135
+ save_dataset: false
136
+ seed: 2020
137
+ seq_len: null
138
+ seq_separator: ' '
139
+ show_progress: false
140
+ shuffle: true
141
+ single_spec: true
142
+ state: INFO
143
+ stopping_step: 10
144
+ test_neg_sample_args:
145
+ distribution: uniform
146
+ sample_num: none
147
+ threshold: null
148
+ topk:
149
+ - 10
150
+ train_batch_size: 4096
151
+ train_neg_sample_args:
152
+ alpha: none
153
+ candidate_num: 0
154
+ distribution: none
155
+ dynamic: false
156
+ sample_num: none
157
+ transform: null
158
+ unload_col: null
159
+ unused_col:
160
+ inter:
161
+ - rating
162
+ use_gpu: true
163
+ user_inter_num_interval: '[10,inf)'
164
+ val_interval:
165
+ rating: '[3,inf)'
166
+ valid_metric: NDCG@10
167
+ valid_metric_bigger: true
168
+ valid_neg_sample_args:
169
+ distribution: uniform
170
+ sample_num: none
171
+ wandb_project: recbole
172
+ weight_decay: 0.0
173
+ worker: 0
174
+
175
+ ```
176
+
177
+ ## Usage