Commit ·
4517db3
1
Parent(s): 8951179
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,41 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
---
|
| 4 |
+
# Fine-Tuned BART Model for Translation
|
| 5 |
+
|
| 6 |
+
This repository contains a fine-tuned version of the BART model for translation tasks. The model is trained to translate human instructions into CLI (Command Line Interface) commands.
|
| 7 |
+
|
| 8 |
+
## Model Details
|
| 9 |
+
|
| 10 |
+
- Model: BART
|
| 11 |
+
- Architecture: `facebook/bart-large`
|
| 12 |
+
- Fine-Tuned on: Custom translation dataset (details in `train_model.py`)
|
| 13 |
+
- Max Input Length: 128 tokens
|
| 14 |
+
- Max Output Length: 50 tokens
|
| 15 |
+
- Beam Search: 4 beams
|
| 16 |
+
|
| 17 |
+
## Usage
|
| 18 |
+
|
| 19 |
+
You can use this model to generate CLI commands from human instructions.
|
| 20 |
+
You can either directly use the model for inference or integrate it into applications
|
| 21 |
+
using the provided Gradio interface.
|
| 22 |
+
|
| 23 |
+
### Inference
|
| 24 |
+
To perform inference using the model, you can load it.
|
| 25 |
+
|
| 26 |
+
### Gradio Interface
|
| 27 |
+
|
| 28 |
+
An interactive Gradio interface is provided for easy model interaction.
|
| 29 |
+
You can run the interface by executing the code in `gradio_app.py`.
|
| 30 |
+
The interface allows you to enter human instructions and get corresponding generated CLI commands.
|
| 31 |
+
|
| 32 |
+
## Model Files
|
| 33 |
+
|
| 34 |
+
- `config.json`: Model configuration file
|
| 35 |
+
- `pytorch_model.bin`: Model weights
|
| 36 |
+
- `tokenizer.json`: Tokenizer configuration file
|
| 37 |
+
- `vocab.txt`: Vocabulary file
|
| 38 |
+
|
| 39 |
+
## Acknowledgments
|
| 40 |
+
|
| 41 |
+
The initial BART model and tokenizer are from the Hugging Face Transformers library (`facebook/bart-large`).
|