3v324v23 commited on
Commit
f3caf97
1 Parent(s): e24dcd1

simple fix of columns

Browse files
Files changed (1) hide show
  1. src/display/about.py +15 -0
src/display/about.py CHANGED
@@ -4,6 +4,21 @@ TITLE = """<h1 align="center" id="space-title">OPEN-MOE-LLM-LEADERBOARD</h1>"""
4
 
5
  INTRODUCTION_TEXT = """
6
  The OPEN-MOE-LLM-LEADERBOARD is specifically designed to assess the performance and efficiency of various Mixture of Experts (MoE) Large Language Models (LLMs). This initiative, driven by the open-source community, aims to comprehensively evaluate these advanced MoE LLMs. We extend our gratitude to the Huggingface for the GPU community grant that supported the initial debugging process, and to [NetMind.AI](https://netmind.ai/home) for their generous GPU donation, which ensures the continuous operation of the Leaderboard.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7
  """
8
  LLM_BENCHMARKS_TEXT = f"""
9
 
 
4
 
5
  INTRODUCTION_TEXT = """
6
  The OPEN-MOE-LLM-LEADERBOARD is specifically designed to assess the performance and efficiency of various Mixture of Experts (MoE) Large Language Models (LLMs). This initiative, driven by the open-source community, aims to comprehensively evaluate these advanced MoE LLMs. We extend our gratitude to the Huggingface for the GPU community grant that supported the initial debugging process, and to [NetMind.AI](https://netmind.ai/home) for their generous GPU donation, which ensures the continuous operation of the Leaderboard.
7
+
8
+ The OPEN-MOE-LLM-LEADERBOARD includes generation and multiple choice tasks to measure the performance and efficiency of MOE LLMs.
9
+
10
+
11
+ Tasks:
12
+ - **Generation Self-consistancy** -- [SelfCheckGPT](https://github.com/potsawee/selfcheckgpt)
13
+ - **Multiple Choice Performance** -- [MMLU](https://arxiv.org/abs/2009.03300)
14
+
15
+ Columns and Metrics:
16
+ - Method: The MOE LLMs inference framework.
17
+ - E2E(s): Average End to End generation time in seconds.
18
+ - PRE(s): Prefilling Time of input prompt in seconds.
19
+ - T/s: Tokens throughout per second.
20
+ - Precision: The precison of used model.
21
+
22
  """
23
  LLM_BENCHMARKS_TEXT = f"""
24