Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
thumbnail: https://huggingface.co/Norod78/gpt-fluentui-flat-svg/raw/main/train/sample16.svg
|
| 3 |
+
license: mit
|
| 4 |
+
library_name: transformers
|
| 5 |
+
pipeline_tag: text-generation
|
| 6 |
+
---
|
| 7 |
+
# gpt-fluentui-flat-svg
|
| 8 |
+
A custom GPT model which was trained upon svg files.
|
| 9 |
+
Specifically the flat emoji variants from [Microsoft's FluentUI repo](https://github.com/microsoft/fluentui-emoji).
|
| 10 |
+
These svn file only consist of "stand-alone" path elements which should make it simpler to train upon and sample from.
|
| 11 |
+
|
| 12 |
+
# training and dataset
|
| 13 |
+
Both Tokenizer and Model were trained using [aitextgen](https://docs.aitextgen.io/)
|
| 14 |
+
The python file which was used for training, the .txt file dataset and a few generated samples can be found [here](https://huggingface.co/Norod78/gpt-fluentui-flat-svg/tree/main/train)
|
| 15 |
+
|
| 16 |
+
# post processing and extracting .svg files from generated samples
|
| 17 |
+
|
| 18 |
+
```
|
| 19 |
+
# Extract from generated output and into a seperate .svg file all sequences which starts with <svg and ends with:
|
| 20 |
+
# A. </svg>
|
| 21 |
+
# B. If the sequence does not end with </svg> then find the last > in the sequence and append </svg> to it
|
| 22 |
+
```
|
| 23 |
+
|
| 24 |
+
# generated samples
|
| 25 |
+
The generated samples below were also created with [this script](https://huggingface.co/Norod78/gpt-fluentui-flat-svg/blob/main/train/atg_train.py)
|
| 26 |
+
|
| 27 |
+
|
| 28 |
+
|