Adding Evaluation Results
#3
by
leaderboard-pr-bot
- opened
README.md
CHANGED
@@ -1,3 +1,108 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
# Firefly-LLaMA2-Chinese: 开源中文LLaMA2大模型
|
2 |
|
3 |
|
@@ -396,3 +501,17 @@ Output:
|
|
396 |
- 模型参数量较小,且未经过人类价值观对其,可能会产生不当言论,不代表本项目观点。
|
397 |
- 与原生中文大模型相比,我们的预训练数据量级还远远不足,且大绝大部分为新闻语料,在一些中文任务上可能表现欠佳。
|
398 |
- 使用本项目的模型时,请遵守对应的官方模型的开源协议。
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
model-index:
|
3 |
+
- name: firefly-llama2-7b-pretrain
|
4 |
+
results:
|
5 |
+
- task:
|
6 |
+
type: text-generation
|
7 |
+
name: Text Generation
|
8 |
+
dataset:
|
9 |
+
name: AI2 Reasoning Challenge (25-Shot)
|
10 |
+
type: ai2_arc
|
11 |
+
config: ARC-Challenge
|
12 |
+
split: test
|
13 |
+
args:
|
14 |
+
num_few_shot: 25
|
15 |
+
metrics:
|
16 |
+
- type: acc_norm
|
17 |
+
value: 48.63
|
18 |
+
name: normalized accuracy
|
19 |
+
source:
|
20 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=YeungNLP/firefly-llama2-7b-pretrain
|
21 |
+
name: Open LLM Leaderboard
|
22 |
+
- task:
|
23 |
+
type: text-generation
|
24 |
+
name: Text Generation
|
25 |
+
dataset:
|
26 |
+
name: HellaSwag (10-Shot)
|
27 |
+
type: hellaswag
|
28 |
+
split: validation
|
29 |
+
args:
|
30 |
+
num_few_shot: 10
|
31 |
+
metrics:
|
32 |
+
- type: acc_norm
|
33 |
+
value: 74.83
|
34 |
+
name: normalized accuracy
|
35 |
+
source:
|
36 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=YeungNLP/firefly-llama2-7b-pretrain
|
37 |
+
name: Open LLM Leaderboard
|
38 |
+
- task:
|
39 |
+
type: text-generation
|
40 |
+
name: Text Generation
|
41 |
+
dataset:
|
42 |
+
name: MMLU (5-Shot)
|
43 |
+
type: cais/mmlu
|
44 |
+
config: all
|
45 |
+
split: test
|
46 |
+
args:
|
47 |
+
num_few_shot: 5
|
48 |
+
metrics:
|
49 |
+
- type: acc
|
50 |
+
value: 41.04
|
51 |
+
name: accuracy
|
52 |
+
source:
|
53 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=YeungNLP/firefly-llama2-7b-pretrain
|
54 |
+
name: Open LLM Leaderboard
|
55 |
+
- task:
|
56 |
+
type: text-generation
|
57 |
+
name: Text Generation
|
58 |
+
dataset:
|
59 |
+
name: TruthfulQA (0-shot)
|
60 |
+
type: truthful_qa
|
61 |
+
config: multiple_choice
|
62 |
+
split: validation
|
63 |
+
args:
|
64 |
+
num_few_shot: 0
|
65 |
+
metrics:
|
66 |
+
- type: mc2
|
67 |
+
value: 39.08
|
68 |
+
source:
|
69 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=YeungNLP/firefly-llama2-7b-pretrain
|
70 |
+
name: Open LLM Leaderboard
|
71 |
+
- task:
|
72 |
+
type: text-generation
|
73 |
+
name: Text Generation
|
74 |
+
dataset:
|
75 |
+
name: Winogrande (5-shot)
|
76 |
+
type: winogrande
|
77 |
+
config: winogrande_xl
|
78 |
+
split: validation
|
79 |
+
args:
|
80 |
+
num_few_shot: 5
|
81 |
+
metrics:
|
82 |
+
- type: acc
|
83 |
+
value: 70.24
|
84 |
+
name: accuracy
|
85 |
+
source:
|
86 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=YeungNLP/firefly-llama2-7b-pretrain
|
87 |
+
name: Open LLM Leaderboard
|
88 |
+
- task:
|
89 |
+
type: text-generation
|
90 |
+
name: Text Generation
|
91 |
+
dataset:
|
92 |
+
name: GSM8k (5-shot)
|
93 |
+
type: gsm8k
|
94 |
+
config: main
|
95 |
+
split: test
|
96 |
+
args:
|
97 |
+
num_few_shot: 5
|
98 |
+
metrics:
|
99 |
+
- type: acc
|
100 |
+
value: 3.26
|
101 |
+
name: accuracy
|
102 |
+
source:
|
103 |
+
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=YeungNLP/firefly-llama2-7b-pretrain
|
104 |
+
name: Open LLM Leaderboard
|
105 |
+
---
|
106 |
# Firefly-LLaMA2-Chinese: 开源中文LLaMA2大模型
|
107 |
|
108 |
|
|
|
501 |
- 模型参数量较小,且未经过人类价值观对其,可能会产生不当言论,不代表本项目观点。
|
502 |
- 与原生中文大模型相比,我们的预训练数据量级还远远不足,且大绝大部分为新闻语料,在一些中文任务上可能表现欠佳。
|
503 |
- 使用本项目的模型时,请遵守对应的官方模型的开源协议。
|
504 |
+
|
505 |
+
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
506 |
+
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_YeungNLP__firefly-llama2-7b-pretrain)
|
507 |
+
|
508 |
+
| Metric |Value|
|
509 |
+
|---------------------------------|----:|
|
510 |
+
|Avg. |46.18|
|
511 |
+
|AI2 Reasoning Challenge (25-Shot)|48.63|
|
512 |
+
|HellaSwag (10-Shot) |74.83|
|
513 |
+
|MMLU (5-Shot) |41.04|
|
514 |
+
|TruthfulQA (0-shot) |39.08|
|
515 |
+
|Winogrande (5-shot) |70.24|
|
516 |
+
|GSM8k (5-shot) | 3.26|
|
517 |
+
|