Skip to content

Commit f8dc941

Browse files
zhangxinyuehfadzhangxinyuehfad
committed
[Doc] Add accuracy report
Signed-off-by: hfadzxy <starmoon_zhang@163.com>
1 parent 724c589 commit f8dc941

File tree

6 files changed

+410
-3
lines changed

6 files changed

+410
-3
lines changed

‎docs/requirements-docs.txt‎

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,4 @@
1-
sphinx
1+
sphinx>=8.2.3
22
sphinx-argparse
33
sphinx-book-theme
44
sphinx-copybutton
@@ -7,3 +7,4 @@ sphinx-togglebutton
77
myst-parser
88
msgspec
99
sphinx-substitution-extensions
10+
snowballstemmer==2.2.0

‎docs/source/developer_guide/evaluation/accuracy_report/Llama-3.1-8B-Instruct.md‎

Lines changed: 163 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,163 @@
1+
# Llama-3.1-8B-Instruct Accuracy Test
2+
<div>
3+
<strong>vLLM version:</strong> vLLM: v0.7.3, vLLM Ascend: v0.7.3 <br>
4+
</div>
5+
<div>
6+
<strong>Software Environment:</strong> CANN: 8.1.0, PyTorch: 2.5.1, torch-npu: 2.5.1 <br>
7+
</div>
8+
<div>
9+
<strong>Hardware Environment</strong>: Atlas A2 Series <br>
10+
</div>
11+
<div>
12+
<strong>Datasets</strong>: ceval-valid,mmlu,gsm8k <br>
13+
</div>
14+
<div>
15+
<strong>Command</strong>:
16+
17+
```bash
18+
export MODEL_AEGS='meta-llama/Llama-3.1-8B-Instruct, max_model_len=4096,dtype=auto,tensor_parallel_size=2,gpu_memory_utilization=0.6'
19+
lm_eval --model vllm --modlel_args $MODEL_ARGS --tasks ceval-valid,gsm8k \
20+
--apply_chat_template --fewshot_as_multiturn --num_fewshot 5 --batch_size 1
21+
```
22+
</div>
23+
<div>&nbsp;</div>
24+
25+
| Task | Filter | n-shot | Metric | Value | Stderr |
26+
|-----------------------|-------:|-------:|----------|--------:|-------:|
27+
| ceval-valid | none | 5 | acc | ↑ 0.5483 | ± 0.0132 |
28+
| mmlu | none | 5 | acc | ↑ 0.6867 | ± 0.0037 |
29+
| gsm8k | flexible-extract | 5 | exact_match | ↑ 0.7286 | ± 0.0122 |
30+
<details>
31+
<summary>ceval-valid</summary>
32+
33+
| Task | Filter | n-shot | Metric | Value | Stderr |
34+
|-----------------------|-------:|-------:|----------|--------:|-------:|
35+
| ceval-valid | none | 5 | acc | ↑ 0.5483 | ± 0.0132 |
36+
| - ceval-valid_accountant | none | 5 | acc | ↑ 0.4898 | ± 0.0722 |
37+
| - ceval-valid_advanced_mathematics | none | 5 | acc | ↑ 0.5263 | ± 0.1177 |
38+
| - ceval-valid_art_studies | none | 5 | acc | ↑ 0.5455 | ± 0.0880 |
39+
| - ceval-valid_basic_medicine | none | 5 | acc | ↑ 0.6842 | ± 0.1096 |
40+
| - ceval-valid_business_administration | none | 5 | acc | ↑ 0.6061 | ± 0.0864 |
41+
| - ceval-valid_chinese_language_and_literature | none | 5 | acc | ↑ 0.4348 | ± 0.1057 |
42+
| - ceval-valid_civil_servant | none | 5 | acc | ↑ 0.4894 | ± 0.0737 |
43+
| - ceval-valid_clinical_medicine | none | 5 | acc | ↑ 0.5455 | ± 0.1087 |
44+
| - ceval-valid_college_chemistry | none | 5 | acc | ↑ 0.4167 | ± 0.1028 |
45+
| - ceval-valid_college_economics | none | 5 | acc | ↑ 0.4545 | ± 0.0678 |
46+
| - ceval-valid_college_physics | none | 5 | acc | ↑ 0.4737 | ± 0.1177 |
47+
| - ceval-valid_college_programming | none | 5 | acc | ↑ 0.5946 | ± 0.0818 |
48+
| - ceval-valid_computer_architecture | none | 5 | acc | ↑ 0.5714 | ± 0.1107 |
49+
| - ceval-valid_computer_network | none | 5 | acc | ↑ 0.7895 | ± 0.0961 |
50+
| - ceval-valid_discrete_mathematics | none | 5 | acc | ↑ 0.4375 | ± 0.1281 |
51+
| - ceval-valid_education_science | none | 5 | acc | ↑ 0.7241 | ± 0.0845 |
52+
| - ceval-valid_electrical_engineer | none | 5 | acc | ↑ 0.4324 | ± 0.0826 |
53+
| - ceval-valid_environmental_impact_assessment_engineer | none | 5 | acc | ↑ 0.5484 | ± 0.0909 |
54+
| - ceval-valid_fire_engineer | none | 5 | acc | ↑ 0.4839 | ± 0.0912 |
55+
| - ceval-valid_high_school_biology | none | 5 | acc | ↑ 0.5263 | ± 0.1177 |
56+
| - ceval-valid_high_school_chemistry | none | 5 | acc | ↑ 0.4737 | ± 0.1177 |
57+
| - ceval-valid_high_school_chinese | none | 5 | acc | ↑ 0.2105 | ± 0.0961 |
58+
| - ceval-valid_high_school_geography | none | 5 | acc | ↑ 0.6842 | ± 0.1096 |
59+
| - ceval-valid_high_school_history | none | 5 | acc | ↑ 0.6500 | ± 0.1094 |
60+
| - ceval-valid_high_school_mathematics | none | 5 | acc | ↑ 0.0000 | ± 0.0000 |
61+
| - ceval-valid_high_school_physics | none | 5 | acc | ↑ 0.3158 | ± 0.1096 |
62+
| - ceval-valid_high_school_politics | none | 5 | acc | ↑ 0.5789 | ± 0.1164 |
63+
| - ceval-valid_ideological_and_moral_cultivation | none | 5 | acc | ↑ 0.8947 | ± 0.0723 |
64+
| - ceval-valid_law | none | 5 | acc | ↑ 0.4583 | ± 0.1039 |
65+
| - ceval-valid_legal_professional | none | 5 | acc | ↑ 0.3913 | ± 0.1041 |
66+
| - ceval-valid_logic | none | 5 | acc | ↑ 0.5000 | ± 0.1091 |
67+
| - ceval-valid_mao_zedong_thought | none | 5 | acc | ↑ 0.5000 | ± 0.1043 |
68+
| - ceval-valid_marxism | none | 5 | acc | ↑ 0.6842 | ± 0.1096 |
69+
| - ceval-valid_metrology_engineer | none | 5 | acc | ↑ 0.5833 | ± 0.1028 |
70+
| - ceval-valid_middle_school_biology | none | 5 | acc | ↑ 0.7143 | ± 0.1010 |
71+
| - ceval-valid_middle_school_chemistry | none | 5 | acc | ↑ 0.8500 | ± 0.0819 |
72+
| - ceval-valid_middle_school_geography | none | 5 | acc | ↑ 0.5833 | ± 0.1486 |
73+
| - ceval-valid_middle_school_history | none | 5 | acc | ↑ 0.5455 | ± 0.1087 |
74+
| - ceval-valid_middle_school_mathematics | none | 5 | acc | ↑ 0.3684 | ± 0.1137 |
75+
| - ceval-valid_middle_school_physics | none | 5 | acc | ↑ 0.6316 | ± 0.1137 |
76+
| - ceval-valid_middle_school_politics | none | 5 | acc | ↑ 0.8095 | ± 0.0878 |
77+
| - ceval-valid_modern_chinese_history | none | 5 | acc | ↑ 0.5217 | ± 0.1065 |
78+
| - ceval-valid_operating_system | none | 5 | acc | ↑ 0.6316 | ± 0.1137 |
79+
| - ceval-valid_physician | none | 5 | acc | ↑ 0.5918 | ± 0.0709 |
80+
| - ceval-valid_plant_protection | none | 5 | acc | ↑ 0.7727 | ± 0.0914 |
81+
| - ceval-valid_probability_and_statistics | none | 5 | acc | ↑ 0.3889 | ± 0.1182 |
82+
| - ceval-valid_professional_tour_guide | none | 5 | acc | ↑ 0.6207 | ± 0.0917 |
83+
| - ceval-valid_sports_science | none | 5 | acc | ↑ 0.6316 | ± 0.1137 |
84+
| - ceval-valid_tax_accountant | none | 5 | acc | ↑ 0.3878 | ± 0.0703 |
85+
| - ceval-valid_teacher_qualification | none | 5 | acc | ↑ 0.7955 | ± 0.0615 |
86+
| - ceval-valid_urban_and_rural_planner | none | 5 | acc | ↑ 0.5217 | ± 0.0745 |
87+
| - ceval-valid_veterinary_medicine | none | 5 | acc | ↑ 0.6087 | ± 0.1041 |
88+
</details>
89+
<details>
90+
<summary>mmlu</summary>
91+
92+
| Task | Filter | n-shot | Metric | Value | Stderr |
93+
|-----------------------|-------:|-------:|----------|--------:|-------:|
94+
| mmlu | none | 5 | acc | ↑ 0.6867 | ± 0.0037 |
95+
| - humanities | none | 5 | acc | ↑ 0.6495 | ± 0.0067 |
96+
| - formal_logic | none | 5 | acc | ↑ 0.5714 | ± 0.0443 |
97+
| - high_school_european_history | none | 5 | acc | ↑ 0.7636 | ± 0.0332 |
98+
| - high_school_us_history | none | 5 | acc | ↑ 0.8186 | ± 0.0270 |
99+
| - high_school_world_history | none | 5 | acc | ↑ 0.8439 | ± 0.0236 |
100+
| - international_law | none | 5 | acc | ↑ 0.8347 | ± 0.0339 |
101+
| - jurisprudence | none | 5 | acc | ↑ 0.7778 | ± 0.0402 |
102+
| - logical_fallacies | none | 5 | acc | ↑ 0.8098 | ± 0.0308 |
103+
| - moral_disputes | none | 5 | acc | ↑ 0.7630 | ± 0.0229 |
104+
| - moral_scenarios | none | 5 | acc | ↑ 0.5687 | ± 0.0166 |
105+
| - philosophy | none | 5 | acc | ↑ 0.7363 | ± 0.0250 |
106+
| - prehistory | none | 5 | acc | ↑ 0.7562 | ± 0.0239 |
107+
| - professional_law | none | 5 | acc | ↑ 0.5111 | ± 0.0128 |
108+
| - world_religions | none | 5 | acc | ↑ 0.8363 | ± 0.0284 |
109+
| - other | none | 5 | acc | ↑ 0.7448 | ± 0.0075 |
110+
| - business_ethics | none | 5 | acc | ↑ 0.7200 | ± 0.0451 |
111+
| - clinical_knowledge | none | 5 | acc | ↑ 0.7509 | ± 0.0266 |
112+
| - college_medicine | none | 5 | acc | ↑ 0.6821 | ± 0.0355 |
113+
| - global_facts | none | 5 | acc | ↑ 0.3900 | ± 0.0490 |
114+
| - human_aging | none | 5 | acc | ↑ 0.6951 | ± 0.0309 |
115+
| - management | none | 5 | acc | ↑ 0.8155 | ± 0.0384 |
116+
| - marketing | none | 5 | acc | ↑ 0.8974 | ± 0.0199 |
117+
| - medical_genetics | none | 5 | acc | ↑ 0.8200 | ± 0.0386 |
118+
| - miscellaneous | none | 5 | acc | ↑ 0.8378 | ± 0.0132 |
119+
| - nutrition | none | 5 | acc | ↑ 0.8039 | ± 0.0227 |
120+
| - professional_accounting | none | 5 | acc | ↑ 0.5532 | ± 0.0297 |
121+
| - professional_medicine | none | 5 | acc | ↑ 0.7721 | ± 0.0255 |
122+
| - virology | none | 5 | acc | ↑ 0.5241 | ± 0.0389 |
123+
| - social sciences | none | 5 | acc | ↑ 0.7797 | ± 0.0073 |
124+
| - econometrics | none | 5 | acc | ↑ 0.6053 | ± 0.0460 |
125+
| - high_school_geography | none | 5 | acc | ↑ 0.8485 | ± 0.0255 |
126+
| - high_school_government_and_politics | none | 5 | acc | ↑ 0.9171 | ± 0.0199 |
127+
| - high_school_macroeconomics | none | 5 | acc | ↑ 0.6923 | ± 0.0234 |
128+
| - high_school_microeconomics | none | 5 | acc | ↑ 0.7647 | ± 0.0276 |
129+
| - high_school_psychology | none | 5 | acc | ↑ 0.8697 | ± 0.0144 |
130+
| - human_sexuality | none | 5 | acc | ↑ 0.8015 | ± 0.0350 |
131+
| - professional_psychology | none | 5 | acc | ↑ 0.7271 | ± 0.0180 |
132+
| - public_relations | none | 5 | acc | ↑ 0.6818 | ± 0.0446 |
133+
| - security_studies | none | 5 | acc | ↑ 0.7224 | ± 0.0287 |
134+
| - sociology | none | 5 | acc | ↑ 0.8358 | ± 0.0262 |
135+
| - us_foreign_policy | none | 5 | acc | ↑ 0.8900 | ± 0.0314 |
136+
| - stem | none | 5 | acc | ↑ 0.5940 | ± 0.0084 |
137+
| - abstract_algebra | none | 5 | acc | ↑ 0.3900 | ± 0.0490 |
138+
| - anatomy | none | 5 | acc | ↑ 0.6741 | ± 0.0405 |
139+
| - astronomy | none | 5 | acc | ↑ 0.7566 | ± 0.0349 |
140+
| - college_biology | none | 5 | acc | ↑ 0.8264 | ± 0.0317 |
141+
| - college_chemistry | none | 5 | acc | ↑ 0.4700 | ± 0.0502 |
142+
| - college_computer_science | none | 5 | acc | ↑ 0.5400 | ± 0.0501 |
143+
| - college_mathematics | none | 5 | acc | ↑ 0.3900 | ± 0.0490 |
144+
| - college_physics | none | 5 | acc | ↑ 0.4314 | ± 0.0493 |
145+
| - computer_security | none | 5 | acc | ↑ 0.8000 | ± 0.0402 |
146+
| - conceptual_physics | none | 5 | acc | ↑ 0.6170 | ± 0.0318 |
147+
| - electrical_engineering | none | 5 | acc | ↑ 0.6552 | ± 0.0396 |
148+
| - elementary_mathematics | none | 5 | acc | ↑ 0.4735 | ± 0.0257 |
149+
| - high_school_biology | none | 5 | acc | ↑ 0.8097 | ± 0.0223 |
150+
| - high_school_chemistry | none | 5 | acc | ↑ 0.6207 | ± 0.0341 |
151+
| - high_school_computer_science | none | 5 | acc | ↑ 0.7300 | ± 0.0446 |
152+
| - high_school_mathematics | none | 5 | acc | ↑ 0.4222 | ± 0.0301 |
153+
| - high_school_physics | none | 5 | acc | ↑ 0.4636 | ± 0.0407 |
154+
| - high_school_statistics | none | 5 | acc | ↑ 0.6065 | ± 0.0333 |
155+
| - machine_learning | none | 5 | acc | ↑ 0.5446 | ± 0.0473 |
156+
</details>
157+
<details>
158+
<summary>gsm8k</summary>
159+
160+
| Task | Filter | n-shot | Metric | Value | Stderr |
161+
|-----------------------|-------:|-------:|----------|--------:|-------:|
162+
| gsm8k | flexible-extract | 5 | exact_match | ↑ 0.7286 | ± 0.0122 |
163+
</details>

0 commit comments

Comments
 (0)