|
4 | 4 |
|
5 | 5 | <!-- [TOC] placeholder - DO NOT REMOVE --> |
6 | 6 |
|
| 7 | +Looking for help? Submit bugs, ask questions, suggest content, and request |
| 8 | +features on our |
| 9 | +[Github issues list](https://github.com/pair-code/lit/issues/). |
| 10 | + |
| 11 | +## Model and Data Types |
| 12 | + |
| 13 | +LIT can handle a variety of models with different input and output types, and |
| 14 | +works with any modern ML framework. For more information, see |
| 15 | +[Framework & Model Support](components.md#framework-and-model-support). |
| 16 | + |
| 17 | +In addition to text, LIT has good support for different modalities, including |
| 18 | +images and tabular data. For examples, see: |
| 19 | + |
| 20 | +* [Image demo](https://github.com/PAIR-code/lit/blob/main/lit_nlp/examples/image_demo.py) - |
| 21 | + image classification, using a Mobilenet model. |
| 22 | +* [Tabular demo](https://github.com/PAIR-code/lit/blob/main/lit_nlp/examples/penguin_demo.py) - |
| 23 | + mult-class classification on tabular (numeric and categorical string) data, |
| 24 | + using the |
| 25 | + [Palmer Penguins](https://www.tensorflow.org/datasets/catalog/penguins) |
| 26 | + dataset. |
| 27 | + |
| 28 | +For more details, see |
| 29 | +[the features guide to input and output types](api.md#type-system). |
| 30 | + |
| 31 | +## Languages |
| 32 | + |
| 33 | +All strings in LIT are unicode and most components use model-provided |
| 34 | +tokenization if available, so in most cases non-English languages and non-Latin |
| 35 | +scripts should work without any modifications. For examples, see: |
| 36 | + |
| 37 | +* [XNLI demo](https://github.com/PAIR-code/lit/blob/main/lit_nlp/examples/xnli_demo.py) - |
| 38 | + cross-lingual NLI, with up to 15 languages supported via a multilingual BERT |
| 39 | + model. |
| 40 | +* [T5 demo](https://github.com/PAIR-code/lit/blob/main/lit_nlp/examples/t5_demo.py) - |
| 41 | + includes WMT data for machine translation |
| 42 | + |
7 | 43 | ## Scale |
8 | 44 |
|
9 | 45 | ### Dataset Size |
@@ -66,41 +102,19 @@ LIT has the ability to create or edit datapoints in the UI and then save them to |
66 | 102 | disk. If you do not want the tool to to be able to write edited datapoints to |
67 | 103 | disk, then pass the `--demo_mode` runtime flag to the LIT server. |
68 | 104 |
|
69 | | -### Managing Access |
| 105 | +### I have proprietary data. Is LIT secure for my team to use? |
70 | 106 |
|
71 | | -The default LIT development server does not implement any explicit access |
72 | | -controls. However, this is just a thin convenience wrapper, and the underlying |
73 | | -WSGI App can be easily exported and used with additional middleware layers or |
74 | | -external serving frameworks. See |
75 | | -[Running LIT in a Docker container](./docker.md) for an example of this usage. |
76 | | - |
77 | | -## Languages |
78 | | - |
79 | | -All strings in LIT are unicode and most components use model-provided |
80 | | -tokenization if available, so in most cases non-English languages and non-Latin |
81 | | -scripts should work without any modifications. For examples, see: |
82 | | - |
83 | | -* [XNLI demo](https://github.com/PAIR-code/lit/blob/main/lit_nlp/examples/xnli_demo.py) - |
84 | | - cross-lingual NLI, with up to 15 languages supported via a multilingual BERT |
85 | | - model. |
86 | | -* [T5 demo](https://github.com/PAIR-code/lit/blob/main/lit_nlp/examples/t5_demo.py) - |
87 | | - includes WMT data for machine translation |
| 107 | +We don't store, collect or share datasets, models or any other information |
| 108 | +loaded into LIT. When you run a LIT server, anyone with access to the web |
| 109 | +address of the server will be able to see data from the loaded datasets and |
| 110 | +interact with the loaded models. If you need to restrict access to a LIT |
| 111 | +server, then make sure to configure the hosting of your LIT server to do so. |
88 | 112 |
|
89 | | -## Data Types |
90 | | - |
91 | | -In addition to text, LIT has good support for different input and output |
92 | | -modalities, including images and tabular data. For examples, see: |
93 | | - |
94 | | -* [Image demo](https://github.com/PAIR-code/lit/blob/main/lit_nlp/examples/image_demo.py) - |
95 | | - image classification, using a Mobilenet model. |
96 | | -* [Tabular demo](https://github.com/PAIR-code/lit/blob/main/lit_nlp/examples/penguin_demo.py) - |
97 | | - mult-class classification on tabular (numeric and categorical string) data, |
98 | | - using the |
99 | | - [Palmer Penguins](https://www.tensorflow.org/datasets/catalog/penguins) |
100 | | - dataset. |
101 | | - |
102 | | -For more details, see |
103 | | -[the features guide to input and output types](api.md#type-system). |
| 113 | +The default LIT development server does not implement any explicit access |
| 114 | +controls. However, this server is just a thin convenience wrapper, and the |
| 115 | +underlying WSGI App can be easily exported and used with additional middleware |
| 116 | +layers or external serving frameworks. See |
| 117 | +[Running LIT in a Docker container](./docker.md) for an example. |
104 | 118 |
|
105 | 119 | ## Workflow and Integrations |
106 | 120 |
|
|
0 commit comments