database-fork/RedPajama.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: RedPajama
    link: https://huggingface.co/togethercomputer/RedPajama-INCITE-7B-Chat
    type: text
    performanceclass: limited
    basemodelname: RedPajama-INCITE-7B-Base
    endmodelname: RedPajama-INCITE-7B-Chat
    endmodellicense: Apache-2.0
    releasedate: 2023-03
    notes: Open AI model developed as a collaboration between various open-source entities.

org:
    name: Together Computer
    link: https://together.ai/
    notes: Together Computer, a cloud platform for generative AI.

# availability:
datasources_basemodel:
    class: open
    link: https://huggingface.co/datasets/togethercomputer/RedPajama-Data-1T
    notes: RedPajama-Data-1T made available on HuggingFace

datasources_endmodel:
    class: open
    link: ["https://huggingface.co/datasets/togethercomputer/RedPajama-Data-Instruct", "https://huggingface.co/datasets/databricks/databricks-dolly-15k", "https://huggingface.co/datasets/OpenAssistant/oasst1"]
    notes: The model was trained on a large collection of diverse data, including Chain-of-Thought (CoT), Public Pool of Prompts (P3) dataset, Natural-Instructions (NI) dataset. Chat-tuning using Databricks-Dolly and OASST1.

weights_basemodel:
    class: open
    link: https://huggingface.co/togethercomputer/RedPajama-INCITE-7B-Base
    notes: Base is RedPajama-INCITE-7B-Base

weights_endmodel:
    class: open
    link: https://huggingface.co/togethercomputer/RedPajama-INCITE-7B-Chat
    notes: Instruction-tuned version made available in parallel with base version.

trainingcode:
    class: partial
    link: https://github.com/togethercomputer/redpajama.cpp/tree/master/examples/redpajama
    notes: Code for datasets made available in exemplary ways; code for training and tuning harder to find.

# documentation:
code:
    class: partial
    link: https://github.com/togethercomputer/redpajama.cpp/tree/master/examples/redpajama
    notes: Code for base LLM and instruction tuning datasets beautifully documented; code specifying training and fine-tuning sparsely documented.

hardware_architecture:
    class: partial
    link: https://together.ai/blog/redpajama
    notes: Architecture detailed on model card, crucial parts appear to be forked from GPT-NeoX

preprint:
    class: closed
    link:
    notes: No preprint found.

paper:
    class: closed
    link:
    notes: No paper found.

modelcard:
    class: open
    link: https://huggingface.co/togethercomputer/RedPajama-INCITE-7B-Chat
    notes: Model card and readme provide details on datasets and training procedure.

datasheet:
    class: open
    link: ["https://huggingface.co/datasets/togethercomputer/RedPajama-Data-1T", "https://huggingface.co/datasets/togethercomputer/RedPajama-Data-Instruct", "https://huggingface.co/datasets/databricks/databricks-dolly-15k", "https://huggingface.co/datasets/OpenAssistant/oasst1"]
    notes: Base data sheet includes links to data and recipes to create from scratch. Other datasets are well-documented.

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

api:
    class: partial
    link: https://huggingface.co/togethercomputer/RedPajama-INCITE-7B-Instruct
    notes: Hosted inference API available through HuggingFace.
    metaprompt: closed

licenses:
    class: partial
    link: https://huggingface.co/togethercomputer/RedPajama-INCITE-7B-Instruct/blob/main/README.md
    notes: Models licensed under Apache 2.0, but note that the data itself is variably licensed and so imposes some limitations.