update README
Browse files
README.md
CHANGED
@@ -14,15 +14,14 @@ size_categories:
|
|
14 |
Datasets accompanying the paper "Pushing the Limits of Pre-training for Time Series Forecasting in the CloudOps Domain".
|
15 |
|
16 |
## Quick Start
|
|
|
|
|
17 |
```python
|
18 |
from datasets import load_dataset
|
19 |
|
20 |
dataset = load_dataset('Salesforce/cloudops_tsf', 'azure_vm_traces_2017')
|
21 |
-
|
22 |
|
23 |
-
## Available Datasets
|
24 |
-
### azure_vm_traces_2017
|
25 |
-
```python
|
26 |
DatasetDict({
|
27 |
train_test: Dataset({
|
28 |
features: ['start', 'target', 'item_id', 'feat_static_cat', 'feat_static_real', 'past_feat_dynamic_real'],
|
@@ -37,6 +36,9 @@ DatasetDict({
|
|
37 |
|
38 |
### borg_cluster_data_2011
|
39 |
```python
|
|
|
|
|
|
|
40 |
DatasetDict({
|
41 |
train_test: Dataset({
|
42 |
features: ['start', 'target', 'item_id', 'feat_static_cat', 'past_feat_dynamic_real'],
|
@@ -51,6 +53,9 @@ DatasetDict({
|
|
51 |
|
52 |
### alibaba_cluster_trace_2018
|
53 |
```python
|
|
|
|
|
|
|
54 |
DatasetDict({
|
55 |
train_test: Dataset({
|
56 |
features: ['start', 'target', 'item_id', 'feat_static_cat', 'past_feat_dynamic_real'],
|
@@ -71,7 +76,7 @@ config = load_dataset_builder('Salesforce/cloudops_tsf', 'azure_vm_traces_2017')
|
|
71 |
print(config)
|
72 |
|
73 |
CloudOpsTSFConfig(
|
74 |
-
name='
|
75 |
version=1.0.0,
|
76 |
data_dir=None,
|
77 |
data_files=None,
|
@@ -79,23 +84,32 @@ CloudOpsTSFConfig(
|
|
79 |
prediction_length=48,
|
80 |
freq='5T',
|
81 |
stride=48,
|
82 |
-
univariate=
|
83 |
-
multivariate=
|
84 |
-
optional_fields=(
|
|
|
|
|
|
|
|
|
85 |
rolling_evaluations=12,
|
86 |
-
test_split_date=Period('
|
87 |
_feat_static_cat_cardinalities={
|
88 |
'pretrain': (
|
89 |
-
('
|
90 |
-
('
|
|
|
|
|
|
|
91 |
'train_test': (
|
92 |
-
('
|
93 |
-
('
|
|
|
|
|
94 |
)
|
95 |
},
|
96 |
-
target_dim=
|
97 |
-
feat_static_real_dim=
|
98 |
-
past_feat_dynamic_real_dim=
|
99 |
)
|
100 |
```
|
101 |
```test_split_date``` is provided to achieve the same train-test split as given in the paper.
|
@@ -117,11 +131,11 @@ The datasets were processed from the following original sources. Please cite the
|
|
117 |
* https://github.com/alibaba/clusterdata
|
118 |
|
119 |
## Citation
|
120 |
-
|
121 |
@article{woo2023pushing,
|
122 |
title={Pushing the Limits of Pre-training for Time Series Forecasting in the CloudOps Domain},
|
123 |
author={Woo, Gerald and Liu, Chenghao and Kumar, Akshat and Sahoo, Doyen},
|
124 |
journal={arXiv preprint arXiv:2310.05063},
|
125 |
year={2023}
|
126 |
}
|
127 |
-
|
|
|
14 |
Datasets accompanying the paper "Pushing the Limits of Pre-training for Time Series Forecasting in the CloudOps Domain".
|
15 |
|
16 |
## Quick Start
|
17 |
+
|
18 |
+
### azure_vm_traces_2017
|
19 |
```python
|
20 |
from datasets import load_dataset
|
21 |
|
22 |
dataset = load_dataset('Salesforce/cloudops_tsf', 'azure_vm_traces_2017')
|
23 |
+
print(dataset)
|
24 |
|
|
|
|
|
|
|
25 |
DatasetDict({
|
26 |
train_test: Dataset({
|
27 |
features: ['start', 'target', 'item_id', 'feat_static_cat', 'feat_static_real', 'past_feat_dynamic_real'],
|
|
|
36 |
|
37 |
### borg_cluster_data_2011
|
38 |
```python
|
39 |
+
dataset = load_dataset('Salesforce/cloudops_tsf', 'borg_cluster_data_2011')
|
40 |
+
print(dataset)
|
41 |
+
|
42 |
DatasetDict({
|
43 |
train_test: Dataset({
|
44 |
features: ['start', 'target', 'item_id', 'feat_static_cat', 'past_feat_dynamic_real'],
|
|
|
53 |
|
54 |
### alibaba_cluster_trace_2018
|
55 |
```python
|
56 |
+
dataset = load_dataset('Salesforce/cloudops_tsf', 'alibaba_cluster_trace_2018')
|
57 |
+
print(dataset)
|
58 |
+
|
59 |
DatasetDict({
|
60 |
train_test: Dataset({
|
61 |
features: ['start', 'target', 'item_id', 'feat_static_cat', 'past_feat_dynamic_real'],
|
|
|
76 |
print(config)
|
77 |
|
78 |
CloudOpsTSFConfig(
|
79 |
+
name='azure_vm_traces_2017',
|
80 |
version=1.0.0,
|
81 |
data_dir=None,
|
82 |
data_files=None,
|
|
|
84 |
prediction_length=48,
|
85 |
freq='5T',
|
86 |
stride=48,
|
87 |
+
univariate=True,
|
88 |
+
multivariate=False,
|
89 |
+
optional_fields=(
|
90 |
+
'feat_static_cat',
|
91 |
+
'feat_static_real',
|
92 |
+
'past_feat_dynamic_real'
|
93 |
+
),
|
94 |
rolling_evaluations=12,
|
95 |
+
test_split_date=Period('2016-12-13 15:55', '5T'),
|
96 |
_feat_static_cat_cardinalities={
|
97 |
'pretrain': (
|
98 |
+
('vm_id', 177040),
|
99 |
+
('subscription_id', 5514),
|
100 |
+
('deployment_id', 15208),
|
101 |
+
('vm_category', 3)
|
102 |
+
),
|
103 |
'train_test': (
|
104 |
+
('vm_id', 17568),
|
105 |
+
('subscription_id', 2713),
|
106 |
+
('deployment_id', 3255),
|
107 |
+
('vm_category', 3)
|
108 |
)
|
109 |
},
|
110 |
+
target_dim=1,
|
111 |
+
feat_static_real_dim=3,
|
112 |
+
past_feat_dynamic_real_dim=2
|
113 |
)
|
114 |
```
|
115 |
```test_split_date``` is provided to achieve the same train-test split as given in the paper.
|
|
|
131 |
* https://github.com/alibaba/clusterdata
|
132 |
|
133 |
## Citation
|
134 |
+
<pre>
|
135 |
@article{woo2023pushing,
|
136 |
title={Pushing the Limits of Pre-training for Time Series Forecasting in the CloudOps Domain},
|
137 |
author={Woo, Gerald and Liu, Chenghao and Kumar, Akshat and Sahoo, Doyen},
|
138 |
journal={arXiv preprint arXiv:2310.05063},
|
139 |
year={2023}
|
140 |
}
|
141 |
+
</pre>
|