Datasets:

Languages:
Portuguese
ArXiv:
Tags:
License:
rishiraj commited on
Commit
e58107e
1 Parent(s): dcfca78

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +66 -0
README.md CHANGED
@@ -31,4 +31,70 @@ configs:
31
  path: data/train-*
32
  - split: test
33
  path: data/test-*
 
 
 
 
 
 
 
34
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
31
  path: data/train-*
32
  - split: test
33
  path: data/test-*
34
+ task_categories:
35
+ - conversational
36
+ - text-generation
37
+ language:
38
+ - pt
39
+ pretty_name: Portuguese Chat
40
+ license: cc-by-nc-4.0
41
  ---
42
+ # Dataset Card for Portuguese Chat
43
+
44
+ We know that current English-first LLMs don’t work well for many other languages, both in terms of performance, latency, and speed. Building instruction datasets for non-English languages is an important challenge that needs to be solved.
45
+
46
+ Dedicated towards addressing this problem, I release 3 new datasets [rishiraj/portuguesechat](https://huggingface.co/datasets/rishiraj/portuguesechat/), [rishiraj/bengalichat](https://huggingface.co/datasets/rishiraj/bengalichat/) & [rishiraj/hindichat](https://huggingface.co/datasets/rishiraj/hindichat/) of 10,000 instructions and demonstrations each. This data can be used for supervised fine-tuning (SFT) to make language multilingual models follow instructions better.
47
+
48
+ ### Dataset Summary
49
+
50
+ [rishiraj/portuguesechat](https://huggingface.co/datasets/rishiraj/portuguesechat/) was modelled after the instruction dataset described in OpenAI's [InstructGPT paper](https://huggingface.co/papers/2203.02155), and is translated from [HuggingFaceH4/no_robots](https://huggingface.co/datasets/HuggingFaceH4/no_robots/) which comprised mostly of single-turn instructions across the following categories:
51
+
52
+ | Category | Count |
53
+ |:-----------|--------:|
54
+ | Generation | 4560 |
55
+ | Open QA | 1240 |
56
+ | Brainstorm | 1120 |
57
+ | Chat | 850 |
58
+ | Rewrite | 660 |
59
+ | Summarize | 420 |
60
+ | Coding | 350 |
61
+ | Classify | 350 |
62
+ | Closed QA | 260 |
63
+ | Extract | 190 |
64
+
65
+ ### Languages
66
+
67
+ The data in [rishiraj/portuguesechat](https://huggingface.co/datasets/rishiraj/portuguesechat/) are in Portuguese (BCP-47 pt).
68
+
69
+ ### Data Fields
70
+
71
+ The data fields are as follows:
72
+
73
+ * `prompt`: Describes the task the model should perform.
74
+ * `prompt_id`: A unique ID for the prompt.
75
+ * `messages`: An array of messages, where each message indicates the role (system, user, assistant) and the content.
76
+ * `category`: Which category the example belongs to (e.g. `Chat` or `Coding`).
77
+ * `text`: Content of `messages` in a format that is compatible with dataset_text_field of SFTTrainer.
78
+
79
+ ### Data Splits
80
+
81
+ | | train_sft | test_sft |
82
+ |---------------|------:| ---: |
83
+ | portuguesechat | 9500 | 500 |
84
+
85
+ ### Licensing Information
86
+
87
+ The dataset is available under the [Creative Commons NonCommercial (CC BY-NC 4.0)](https://creativecommons.org/licenses/by-nc/4.0/legalcode).
88
+
89
+ ### Citation Information
90
+
91
+ ```
92
+ @misc{portuguesechat,
93
+ author = {Rishiraj Acharya},
94
+ title = {Portuguese Chat},
95
+ year = {2023},
96
+ publisher = {Hugging Face},
97
+ journal = {Hugging Face repository},
98
+ howpublished = {\url{https://huggingface.co/datasets/rishiraj/portuguesechat}}
99
+ }
100
+ ```