Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Model Catalogue
|
2 |
+
|
3 |
+
- Repository of boutique AI models
|
4 |
+
- Still a work in progress
|
5 |
+
|
6 |
+
## Pretrained base models for embedding
|
7 |
+
|
8 |
+
English models were finetuned on a subset of [Zyphra/Zyda-2](https://huggingface.co/datasets/Zyphra/Zyda-2).
|
9 |
+
- [snowflake-arctic-embed-xs-zyda-2](https://huggingface.co/agentlans/snowflake-arctic-embed-xs-zyda-2)
|
10 |
+
- [deberta-v3-xsmall-zyda-2](https://huggingface.co/agentlans/deberta-v3-xsmall-zyda-2)
|
11 |
+
- [deberta-v3-base-zyda-2](https://huggingface.co/agentlans/deberta-v3-base-zyda-2)
|
12 |
+
|
13 |
+
Multilingual models were aligned using [agentlans/en-translations](https://huggingface.co/datasets/agentlans/en-translations).
|
14 |
+
- [multilingual-e5-small-aligned](https://huggingface.co/agentlans/multilingual-e5-small-aligned)
|
15 |
+
- [distilbert-base-multilingual-cased-aligned](https://huggingface.co/agentlans/distilbert-base-multilingual-cased-aligned)
|
16 |
+
|
17 |
+
## Text statistics models
|
18 |
+
|
19 |
+
- **Input:** text
|
20 |
+
- **Output:** number
|
21 |
+
|
22 |
+
| **Base model** | **Language** | **Quality** | **Readability** | **Sentiment** |
|
23 |
+
|:-----------------------------:|:------------:|:------------------------------------------------------------------------------:|:----------------------------------------------------------------------------------:|:--------------------------------------------------------------------------------:|
|
24 |
+
| deberta-v3-xsmall-zyda-2 | English only | [Link](https://huggingface.co/agentlans/deberta-v3-xsmall-zyda-2-quality) | [Link](https://huggingface.co/agentlans/deberta-v3-xsmall-zyda-2-readability) | [Link](https://huggingface.co/agentlans/deberta-v3-xsmall-zyda-2-sentiment) |
|
25 |
+
| deberta-v3-base-zyda-2 | English only | [Link](https://huggingface.co/agentlans/deberta-v3-base-zyda-2-quality) | [Link](https://huggingface.co/agentlans/deberta-v3-base-zyda-2-readability) | [Link](https://huggingface.co/agentlans/deberta-v3-base-zyda-2-sentiment) |
|
26 |
+
| multilingual-e5-small-aligned | Multilingual | [Link](https://huggingface.co/agentlans/multilingual-e5-small-aligned-quality) | [Link](https://huggingface.co/agentlans/multilingual-e5-small-aligned-readability) | [Link](https://huggingface.co/agentlans/multilingual-e5-small-aligned-sentiment) |
|
27 |
+
| mdeberta-v3-base | Multilingual | [Link](https://huggingface.co/agentlans/mdeberta-v3-base-quality) | [Link](https://huggingface.co/agentlans/mdeberta-v3-base-readability) | [Link](https://huggingface.co/agentlans/mdeberta-v3-base-sentiment) |
|
28 |
+
|
29 |
+
Note: the `mdeberta-v3-base` models were trained on a previous version of the dataset, not the complete dataset
|
30 |
+
|
31 |
+
## Small text-to-text models
|
32 |
+
|
33 |
+
- **Input:** text
|
34 |
+
- **Output:** text
|
35 |
+
|
36 |
+
These are English only.
|
37 |
+
|
38 |
+
| **Task** | **Model** | **Dataset** |
|
39 |
+
|:------------------:|:---------------------------------------------------------------------------------:|:------------------------------------------------------------------------------------------------------:|
|
40 |
+
| Keyword extraction | [flan-t5-small-keywords](https://huggingface.co/agentlans/flan-t5-small-keywords) | [wikipedia-paragraph-keywords](https://huggingface.co/datasets/agentlans/wikipedia-paragraph-keywords) |
|
41 |
+
| Title generation | [flan-t5-small-title](https://huggingface.co/agentlans/flan-t5-small-title) | [wikipedia-paragraph-titles](https://huggingface.co/datasets/agentlans/wikipedia-paragraph-titles) |
|
42 |
+
|
43 |
+
## Natural language inference (NLI) models
|
44 |
+
|
45 |
+
- **Input:** text
|
46 |
+
- **Output:** label (entailment, neutral, or contradiction)
|
47 |
+
|
48 |
+
These are English only.
|
49 |
+
|
50 |
+
- [all-MiniLM-L6-v2-nli](https://huggingface.co/agentlans/all-MiniLM-L6-v2-nli)
|
51 |
+
- [bge-small-en-v1.5-nli](https://huggingface.co/agentlans/bge-small-en-v1.5-nli)
|
52 |
+
- [e5-small-v2-nli](https://huggingface.co/agentlans/e5-small-v2-nli)
|
53 |
+
- [mobilebert-uncased-nli](https://huggingface.co/agentlans/mobilebert-uncased-nli)
|
54 |
+
- [NoInstruct-small-Embedding-v0-nli](https://huggingface.co/agentlans/NoInstruct-small-Embedding-v0-nli)
|
55 |
+
- [snowflake-arctic-embed-s-nli](https://huggingface.co/agentlans/snowflake-arctic-embed-s-nli)
|
56 |
+
- [snowflake-arctic-embed-xs-nli](https://huggingface.co/agentlans/snowflake-arctic-embed-xs-nli)
|
57 |
+
- [TinyBERT_General_4L_312D-nli](https://huggingface.co/agentlans/TinyBERT_General_4L_312D-nli)
|