database-fork/CodeT5.yaml at preview · Language-Technology-Assessment/database-fork · 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
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
---
# Thank you for contributing!
# In filling out this yaml file, please follow the criteria as described here:
# https://osai-index.eu/contribute

# You're free to build on this work and reuse the data. It is licensed under CC-BY 4.0, with the
# stipulation that attribution should come in the form of a link to https://osai-index.eu/
# and a citation to the peer-reviewed paper in which the dataset & criteria were published:

# Liesenfeld, A. and Dingemanse, M., 2024. Rethinking open source generative AI: open-washing and the EU AI Act. In Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency (pp. 1774-1787).

# Organization tags:
# - National origin: United States
# - Contributor type: Non-academic (Company)

system:
    name: CodeT5+
    link: https://huggingface.co/Salesforce/instructcodet5p-16b
    type: code
    performanceclass: full
    basemodelname: CodeGen-16B-NL
    endmodelname: InstructCodeT5p-16B
    endmodellicense: BSD 3-Clause
    releasedate: 2023-05
    notes: T5-based open code model. Model components initialized with CodeGen.

org:
    name: Salesforce
    link: https://www.salesforceairesearch.com/
    notes: Major AI research company.

# availability:
datasources_basemodel:
    class: partial
    link: ["https://pile.eleuther.ai/", "https://cloud.google.com/bigquery/docs/datasets-intro"]
    notes: Model trained on The Pile, BigQuery datasets, and BigQuery. Not all datasets were found.

datasources_endmodel:
    class: open
    link: ["https://huggingface.co/datasets/code-search-net/code_search_net", "https://huggingface.co/datasets/codeparrot/github-code", "https://github.com/sahil280114/codealpaca"]
    notes: Pretraining data from CodeSearchNet and GitHub Code. Instruction-tuning data from CodeAlpaca.

weights_basemodel:
    class: open
    link: https://huggingface.co/Salesforce/codegen-16B-nl
    notes: Model weights made available on HuggingFace.

weights_endmodel:
    class: open
    link: https://huggingface.co/Salesforce/instructcodet5p-16b
    notes: Model weights made available on HuggingFace.

trainingcode:
    class: closed
    link: https://github.com/salesforce/CodeT5/tree/main
    notes: Repo exists, but does not contain training code.

# documentation:
code:
    class: closed
    link:
    notes: No training code, so no documentation.

hardware_architecture:
    class: open
    link: https://arxiv.org/pdf/2305.07922
    notes: Hardware architecture described in preprint.

preprint:
    class: open
    link: https://arxiv.org/pdf/2305.07922
    notes: Preprint available on arXiv.

paper:
    class: open
    link: https://aclanthology.org/2023.emnlp-main.68/
    notes: Paper published as conference paper.

modelcard:
    class: closed
    link: https://huggingface.co/Salesforce/instructcodet5p-16b
    notes: Model card provides the requisite detail.

datasheet:
    class: partial
    link: ["https://pile.eleuther.ai/", "https://cloud.google.com/bigquery/docs/datasets-intro", "https://huggingface.co/datasets/code-search-net/code_search_net", "https://huggingface.co/datasets/codeparrot/github-code", "https://github.com/sahil280114/codealpaca"]
    notes: Data sheets available for most data sources, with varying levels of detail regarding collection and curation procedures.

# access:
package:
    class: closed
    link:
    notes: No package found.

api:
    class: closed
    link:
    notes: No API found.
    metaprompt: closed

licenses:
    class: open
    link: https://huggingface.co/Salesforce/instructcodet5p-16b
    notes: BSD 3-Clause, an open-source license.