Update README.md
Browse files
README.md
CHANGED
@@ -10,37 +10,37 @@ dataset_info:
|
|
10 |
- name: content
|
11 |
dtype: string
|
12 |
splits:
|
13 |
-
- name:
|
14 |
num_bytes: 5567696
|
15 |
num_examples: 3086
|
16 |
-
- name:
|
17 |
num_bytes: 806635
|
18 |
num_examples: 1007
|
19 |
-
- name:
|
20 |
num_bytes: 666517
|
21 |
num_examples: 421
|
22 |
-
- name:
|
23 |
num_bytes: 228494
|
24 |
num_examples: 240
|
25 |
-
- name:
|
26 |
num_bytes: 1068526
|
27 |
num_examples: 458
|
28 |
-
- name:
|
29 |
num_bytes: 27611061
|
30 |
num_examples: 10603
|
31 |
-
- name:
|
32 |
num_bytes: 11069103
|
33 |
num_examples: 1485
|
34 |
-
- name:
|
35 |
num_bytes: 2551884
|
36 |
num_examples: 1508
|
37 |
-
- name:
|
38 |
num_bytes: 13986060
|
39 |
num_examples: 5631
|
40 |
-
- name:
|
41 |
num_bytes: 63555976
|
42 |
num_examples: 24439
|
43 |
-
- name:
|
44 |
num_bytes: 228494
|
45 |
num_examples: 240
|
46 |
download_size: 67916523
|
@@ -48,27 +48,27 @@ dataset_info:
|
|
48 |
configs:
|
49 |
- config_name: default
|
50 |
data_files:
|
51 |
-
- split:
|
52 |
path: data/train_douban_sft-*
|
53 |
-
- split:
|
54 |
path: data/train_human_value_sft-*
|
55 |
-
- split:
|
56 |
path: data/train_logi_qa_sft-*
|
57 |
-
- split:
|
58 |
path: data/train_ruozhiba_sft-*
|
59 |
-
- split:
|
60 |
path: data/train_segmentfault_sft-*
|
61 |
-
- split:
|
62 |
path: data/train_wiki_sft-*
|
63 |
-
- split:
|
64 |
path: data/train_wikihow_sft-*
|
65 |
-
- split:
|
66 |
path: data/train_xhs_sft-*
|
67 |
-
- split:
|
68 |
path: data/train_zhihu_sft-*
|
69 |
-
- split:
|
70 |
path: data/train_sft-*
|
71 |
-
- split:
|
72 |
path: data/test_sft-*
|
73 |
task_categories:
|
74 |
- question-answering
|
@@ -83,7 +83,7 @@ size_categories:
|
|
83 |
数据完全来自于[COIG-CQIA](https://huggingface.co/datasets/m-a-p/COIG-CQIA)。
|
84 |
|
85 |
暂时忽略了chinese_traditional,coig_pc,exam,finance这些转换麻烦或者语义上不适合当QA数据集的subset。
|
86 |
-
其中
|
87 |
|
88 |
经过reformat的本数据集可以直接使用[alignment-handbook](https://github.com/huggingface/alignment-handbook) 进行sft。
|
89 |
|
|
|
10 |
- name: content
|
11 |
dtype: string
|
12 |
splits:
|
13 |
+
- name: train_douban
|
14 |
num_bytes: 5567696
|
15 |
num_examples: 3086
|
16 |
+
- name: train_human_value
|
17 |
num_bytes: 806635
|
18 |
num_examples: 1007
|
19 |
+
- name: train_logi_qa
|
20 |
num_bytes: 666517
|
21 |
num_examples: 421
|
22 |
+
- name: train_ruozhiba
|
23 |
num_bytes: 228494
|
24 |
num_examples: 240
|
25 |
+
- name: train_segmentfault
|
26 |
num_bytes: 1068526
|
27 |
num_examples: 458
|
28 |
+
- name: train_wiki
|
29 |
num_bytes: 27611061
|
30 |
num_examples: 10603
|
31 |
+
- name: train_wikihow
|
32 |
num_bytes: 11069103
|
33 |
num_examples: 1485
|
34 |
+
- name: train_xhs
|
35 |
num_bytes: 2551884
|
36 |
num_examples: 1508
|
37 |
+
- name: train_zhihu
|
38 |
num_bytes: 13986060
|
39 |
num_examples: 5631
|
40 |
+
- name: train
|
41 |
num_bytes: 63555976
|
42 |
num_examples: 24439
|
43 |
+
- name: test
|
44 |
num_bytes: 228494
|
45 |
num_examples: 240
|
46 |
download_size: 67916523
|
|
|
48 |
configs:
|
49 |
- config_name: default
|
50 |
data_files:
|
51 |
+
- split: train_douban
|
52 |
path: data/train_douban_sft-*
|
53 |
+
- split: train_human_value
|
54 |
path: data/train_human_value_sft-*
|
55 |
+
- split: train_logi_qa
|
56 |
path: data/train_logi_qa_sft-*
|
57 |
+
- split: train_ruozhiba
|
58 |
path: data/train_ruozhiba_sft-*
|
59 |
+
- split: train_segmentfault
|
60 |
path: data/train_segmentfault_sft-*
|
61 |
+
- split: train_wiki
|
62 |
path: data/train_wiki_sft-*
|
63 |
+
- split: train_wikihow
|
64 |
path: data/train_wikihow_sft-*
|
65 |
+
- split: train_xhs
|
66 |
path: data/train_xhs_sft-*
|
67 |
+
- split: train_zhihu
|
68 |
path: data/train_zhihu_sft-*
|
69 |
+
- split: train
|
70 |
path: data/train_sft-*
|
71 |
+
- split: test
|
72 |
path: data/test_sft-*
|
73 |
task_categories:
|
74 |
- question-answering
|
|
|
83 |
数据完全来自于[COIG-CQIA](https://huggingface.co/datasets/m-a-p/COIG-CQIA)。
|
84 |
|
85 |
暂时忽略了chinese_traditional,coig_pc,exam,finance这些转换麻烦或者语义上不适合当QA数据集的subset。
|
86 |
+
其中train是全集,test是ruozhiba,以便代码能够跑通。
|
87 |
|
88 |
经过reformat的本数据集可以直接使用[alignment-handbook](https://github.com/huggingface/alignment-handbook) 进行sft。
|
89 |
|