AccelMark/NOTICE at main · ChiangYuhsin/AccelMark · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
AccelMark
Copyright 2024-2026 Juhao Liang and The AccelMark Contributors

This product includes software developed as part of the AccelMark project
(https://github.com/FreedomIntelligence/AccelMark).

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

================================================================================
Third-party bundled data
================================================================================

The AccelMark source tree includes a small amount of third-party data so that
benchmark runs are fully reproducible without network access. Each bundled
dataset retains its upstream license; the Apache 2.0 license above covers only
the AccelMark code, schemas, and configuration around it.

--------------------------------------------------------------------------------
1. datasets/sharegpt_standard_v1/requests.jsonl   (500 prompts)
   datasets/sharegpt_edge_v1/requests.jsonl       (500 prompts)
   datasets/sharegpt_longctx_v1/requests.jsonl    (200 prompts)
--------------------------------------------------------------------------------

  Derived from the ShareGPT GPT-4 conversational dataset curated by:

    shibing624/sharegpt_gpt4
    https://huggingface.co/datasets/shibing624/sharegpt_gpt4
    License: CC BY 4.0
            (https://creativecommons.org/licenses/by/4.0/)

  The upstream corpus was assembled from publicly shared ChatGPT/GPT-4
  conversations. AccelMark's variants are filtered subsets used as fixed
  benchmark inputs; no derivation is intended as the authoritative copy.

  Attribution: shibing624/sharegpt_gpt4 contributors, distributed under CC BY 4.0.

  See datasets/<name>/README.md for the per-subset filtering criteria and
  token statistics.

--------------------------------------------------------------------------------
2. schema/accuracy_subset.jsonl                   (100 multiple-choice items)
--------------------------------------------------------------------------------

  A 100-question subset of MMLU (Massive Multitask Language Understanding):

    Hendrycks, D., Burns, C., Basart, S., Zou, A., Mazeika, M., Song, D.,
    & Steinhardt, J. (2021). "Measuring Massive Multitask Language
    Understanding." International Conference on Learning Representations.
    https://arxiv.org/abs/2009.03300
    https://github.com/hendrycks/test

    License: MIT
            (https://opensource.org/licenses/MIT)

  AccelMark uses this subset purely as an accuracy gate (model-quality
  sanity check) — it is NOT a measurement of MMLU performance. The subset
  is immutable; see CONTRIBUTING.md "A few rules".

================================================================================
Third-party software dependencies
================================================================================

AccelMark's Python runtime dependencies (jsonschema, numpy, pyyaml, …) and
the framework backends invoked by each runner (vLLM, SGLang, mlx-lm,
vllm-ascend, vllm-rocm, vllm-tpu, vllm-musa, …) retain their own licenses.
See each runner's requirements.txt for pinned versions; see the upstream
projects for the corresponding license terms.