Spaces:
Running
Running
kz-transformers
commited on
Update src/display/about.py
Browse files- src/display/about.py +14 -43
src/display/about.py
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
from src.display.utils import ModelType
|
2 |
|
3 |
-
TITLE = """<h1 style="text-align:left;float:left; id="space-title">🤗
|
4 |
|
5 |
INTRODUCTION_TEXT = """
|
6 |
"""
|
@@ -16,18 +16,18 @@ icons = f"""
|
|
16 |
LLM_BENCHMARKS_TEXT = """
|
17 |
## En:
|
18 |
|
19 |
-
|
20 |
-
-
|
21 |
-
-
|
22 |
-
-
|
23 |
-
-
|
24 |
-
-
|
25 |
-
Each task contains
|
26 |
## Instructions for Use
|
27 |
### Installation
|
28 |
To install the necessary library, run the following command:
|
29 |
```bash
|
30 |
-
pip install git+https://github.com/
|
31 |
```
|
32 |
### Execution
|
33 |
To run the benchmark, use the following command:
|
@@ -35,10 +35,10 @@ To run the benchmark, use the following command:
|
|
35 |
!lm_eval \
|
36 |
--model hf \
|
37 |
--model_args pretrained={hf/model},dtype=float16 \
|
38 |
-
--batch_size
|
39 |
--apply_chat_template \
|
40 |
--num_fewshot 0 \
|
41 |
-
--tasks
|
42 |
--output output
|
43 |
```
|
44 |
### Results
|
@@ -47,38 +47,9 @@ After executing the above command, a JSON file will be created in the `output` d
|
|
47 |
If cheating or attempts to modify the output file are detected, we reserve the right to delete your submission.
|
48 |
Thank you for participating!
|
49 |
|
50 |
-
##
|
51 |
-
|
52 |
-
|
53 |
-
- Сложные междисциплинные вопросы (MMLUpro-ru)
|
54 |
-
- Законы Российской Федерации (lawmc)
|
55 |
-
- Популярная музыка (musicmc)
|
56 |
-
- Книги (bookmc)
|
57 |
-
- Фильмы (moviemc)
|
58 |
-
Каждая задача содержит 12 вариантов ответа, mmlupro-ru из 10.
|
59 |
-
## Инструкция по использованию
|
60 |
-
### Установка
|
61 |
-
Для установки необходимой библиотеки выполните следующую команду:
|
62 |
-
```bash
|
63 |
-
pip install git+https://github.com/VikhrModels/lm_eval_mc.git --upgrade --force-reinstall --no-deps
|
64 |
-
```
|
65 |
-
### Запуск
|
66 |
-
Для запуска бенча используйте следующую команду:
|
67 |
-
```bash
|
68 |
-
!lm_eval \
|
69 |
-
--model hf \
|
70 |
-
--model_args pretrained={hf/model},dtype=float16 \
|
71 |
-
--batch_size 8 \
|
72 |
-
--apply_chat_template \
|
73 |
-
--num_fewshot 0 \
|
74 |
-
--tasks musicmc,moviemc,bookmc,lawmc,mmluproru \
|
75 |
-
--output output
|
76 |
-
```
|
77 |
-
### Результаты
|
78 |
-
После выполнения команды выше, в каталоге `output` будет создан файл в формате json, его необходимо прикрепить. Этот файл содержит результаты выполнения задач и описание сессии, его **нельзя модифицировать**.
|
79 |
-
## Политика против читерства
|
80 |
-
При обнаружении читерства или попыток модификации выходного файла, мы оставляем за собой право удалить ваш сабмишен.
|
81 |
-
Спасибо за участие!
|
82 |
|
83 |
Cite: @misc{aleks2024vikhr,
|
84 |
title={Vikhr: The Family of Open-Source Instruction-Tuned Large Language Models for Russian},
|
|
|
1 |
from src.display.utils import ModelType
|
2 |
|
3 |
+
TITLE = """<h1 style="text-align:left;float:left; id="space-title">🤗 Kaz LLM Leaderboard</h1> <h3 style="text-align:left;float:left;> Track, rank and evaluate open LLMs and chatbots </h3>"""
|
4 |
|
5 |
INTRODUCTION_TEXT = """
|
6 |
"""
|
|
|
16 |
LLM_BENCHMARKS_TEXT = """
|
17 |
## En:
|
18 |
|
19 |
+
Kaz LLM is a benchmark for LLM with multiple-choice tasks on the following topics:
|
20 |
+
- mmlu-translated-kk
|
21 |
+
- gsm8k-kk-translated
|
22 |
+
- kazakh-unified-national-testing-mc
|
23 |
+
- kazakh-constitution-mc
|
24 |
+
- kazakh-dastur-mc
|
25 |
+
Each task contains 4 answer choices, mmlu-translated-kk has ??.
|
26 |
## Instructions for Use
|
27 |
### Installation
|
28 |
To install the necessary library, run the following command:
|
29 |
```bash
|
30 |
+
pip install git+https://github.com/horde-research/kaz-llm-eval-lb.git --upgrade --force-reinstall --no-deps
|
31 |
```
|
32 |
### Execution
|
33 |
To run the benchmark, use the following command:
|
|
|
35 |
!lm_eval \
|
36 |
--model hf \
|
37 |
--model_args pretrained={hf/model},dtype=float16 \
|
38 |
+
--batch_size 1 \
|
39 |
--apply_chat_template \
|
40 |
--num_fewshot 0 \
|
41 |
+
--tasks kazakh-dastur-mc \
|
42 |
--output output
|
43 |
```
|
44 |
### Results
|
|
|
47 |
If cheating or attempts to modify the output file are detected, we reserve the right to delete your submission.
|
48 |
Thank you for participating!
|
49 |
|
50 |
+
## KZ:
|
51 |
+
|
52 |
+
to be filled
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
53 |
|
54 |
Cite: @misc{aleks2024vikhr,
|
55 |
title={Vikhr: The Family of Open-Source Instruction-Tuned Large Language Models for Russian},
|