aladar commited on
Commit
c57232d
·
verified ·
1 Parent(s): 7fe0d86

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +75 -27
README.md CHANGED
@@ -1,27 +1,75 @@
1
- ---
2
- dataset_info:
3
- config_name: triplet
4
- features:
5
- - name: query
6
- dtype: string
7
- - name: positive
8
- dtype: string
9
- - name: negative
10
- dtype: string
11
- splits:
12
- - name: train
13
- num_bytes: 12581563.792427007
14
- num_examples: 42076
15
- - name: test
16
- num_bytes: 3149278.207572993
17
- num_examples: 10532
18
- download_size: 1254810
19
- dataset_size: 15730842.0
20
- configs:
21
- - config_name: triplet
22
- data_files:
23
- - split: train
24
- path: triplet/train-*
25
- - split: test
26
- path: triplet/test-*
27
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ dataset_info:
3
+ config_name: triplet
4
+ features:
5
+ - name: query
6
+ dtype: string
7
+ - name: positive
8
+ dtype: string
9
+ - name: negative
10
+ dtype: string
11
+ splits:
12
+ - name: train
13
+ num_bytes: 12581563.792427007
14
+ num_examples: 42076
15
+ - name: test
16
+ num_bytes: 3149278.207572993
17
+ num_examples: 10532
18
+ download_size: 1254810
19
+ dataset_size: 15730842
20
+ configs:
21
+ - config_name: triplet
22
+ data_files:
23
+ - split: train
24
+ path: triplet/train-*
25
+ - split: test
26
+ path: triplet/test-*
27
+ task_categories:
28
+ - sentence-similarity
29
+ ---
30
+
31
+ This dataset is the triplet subset of https://huggingface.co/datasets/sentence-transformers/sql-questions with a train and test split.
32
+
33
+ The test split can be passed to [`TripletEvaluator`](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#tripletevaluator).
34
+
35
+ The train and test spilts don't have any queries in common.
36
+
37
+ <details>
38
+ <summary>Here's the full script used to generate this dataset</summary>
39
+
40
+ ```python
41
+ import os
42
+
43
+ import datasets
44
+ from sklearn.model_selection import train_test_split
45
+
46
+
47
+ dataset = datasets.load_dataset(
48
+ "sentence-transformers/sql-questions", "triplet", split="train"
49
+ )
50
+
51
+ queries_unique = list({record["query"]: None for record in dataset})
52
+ # Use a dict for deterministic (insertion) order
53
+ len(queries_unique)
54
+
55
+ queries_tr, queries_te = train_test_split(
56
+ queries_unique, test_size=0.2, random_state=42
57
+ )
58
+
59
+ queries_tr = set(queries_tr)
60
+ queries_te = set(queries_te)
61
+ train_dataset = dataset.filter(lambda record: record["query"] in queries_tr)
62
+ test_dataset = dataset.filter(lambda record: record["query"] in queries_te)
63
+
64
+ assert not set(train_dataset["query"]) & set(test_dataset["query"])
65
+ assert len(train_dataset) + len(test_dataset) == len(dataset)
66
+
67
+
68
+ dataset_dict = datasets.DatasetDict({"train": train_dataset, "test": test_dataset})
69
+ dataset_dict.push_to_hub(
70
+ "aladar/sql-questions", config_name="triplet", token=os.environ["HF_TOKEN_CREATE"]
71
+ )
72
+
73
+ ```
74
+
75
+ </details>