Update README.md
Browse files
README.md
CHANGED
@@ -1,47 +1,51 @@
|
|
1 |
---
|
2 |
license: mit
|
3 |
tags:
|
4 |
-
-
|
|
|
|
|
5 |
model-index:
|
6 |
- name: sinhala-gpt2
|
7 |
results: []
|
8 |
widget:
|
9 |
-
- text:
|
10 |
-
- text:
|
11 |
-
- text:
|
12 |
-
- text:
|
13 |
-
- text:
|
14 |
inference:
|
15 |
parameters:
|
16 |
do_sample: false
|
17 |
temperature: 0.3
|
|
|
|
|
18 |
---
|
19 |
|
20 |
# sinhala-gpt-lyrics
|
21 |
|
22 |
-
This particular model has undergone fine-tuning based on the [gpt2](https://huggingface.co/gpt2) architecture, utilizing a dataset of
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
|
|
27 |
|
28 |
## Training procedure
|
29 |
-
The model was trained for approximately
|
30 |
|
31 |
## Usage Details
|
32 |
|
33 |
```python
|
34 |
from transformers import AutoTokenizer, AutoModelForCausalLM,pipeline
|
35 |
|
36 |
-
tokenizer = AutoTokenizer.from_pretrained("Ransaka/sinhala-
|
37 |
-
model = AutoModelForCausalLM.from_pretrained("Ransaka/sinhala-
|
38 |
-
generator
|
39 |
-
generator("දුර") #දුර ඈත පාසැල් වියේ.
|
40 |
```
|
41 |
or using git
|
42 |
```bash
|
43 |
git lfs install
|
44 |
-
git clone https://huggingface.co/Ransaka/sinhala-
|
45 |
```
|
46 |
|
47 |
### Training hyperparameters
|
@@ -69,4 +73,4 @@ The following hyperparameters were used during training:
|
|
69 |
- Transformers 4.26.1
|
70 |
- Pytorch 1.13.0
|
71 |
- Datasets 2.1.0
|
72 |
-
- Tokenizers 0.13.2
|
|
|
1 |
---
|
2 |
license: mit
|
3 |
tags:
|
4 |
+
- pytorch
|
5 |
+
- sinhala
|
6 |
+
- gpt2
|
7 |
model-index:
|
8 |
- name: sinhala-gpt2
|
9 |
results: []
|
10 |
widget:
|
11 |
+
- text: මහ
|
12 |
+
- text: සංවිධ
|
13 |
+
- text: දුර්ලභ
|
14 |
+
- text: තනිවීලා
|
15 |
+
- text: ඔබ
|
16 |
inference:
|
17 |
parameters:
|
18 |
do_sample: false
|
19 |
temperature: 0.3
|
20 |
+
language:
|
21 |
+
- si
|
22 |
---
|
23 |
|
24 |
# sinhala-gpt-lyrics
|
25 |
|
26 |
+
This particular model has undergone fine-tuning based on the [gpt2](https://huggingface.co/gpt2) architecture, utilizing a dataset of Sinhala NEWS from various sources.
|
27 |
+
Even though this version of GPT-2 has been finely tuned and is quite simple, it is still capable of generating news articles that are identical. Take, for example, the following samples(Some of them are hilarious though :D):
|
28 |
+
- "ඔබ විසින් මෙම විරෝධතාව සංවිධානය කර තිබුණේ නැහැ කියලා හිටපු ජනාධිපති මහ"
|
29 |
+
- "දුර්ලභ ගණයේ විශ්වවිද්යාල ප්රතිපාදන කොමිෂන් සභාවේ සභාපති මහාචාර්ය ජී එල්"
|
30 |
+
-
|
31 |
+
⚠️ Since the dataset used for this model is mostly composed of news articles, it is heavily biased towards generating news content. This bias may become apparent during the generation process.
|
32 |
|
33 |
## Training procedure
|
34 |
+
The model was trained for approximately 12+ hours on Kaggle GPUs.
|
35 |
|
36 |
## Usage Details
|
37 |
|
38 |
```python
|
39 |
from transformers import AutoTokenizer, AutoModelForCausalLM,pipeline
|
40 |
|
41 |
+
tokenizer = AutoTokenizer.from_pretrained("Ransaka/sinhala-gpt2")
|
42 |
+
model = AutoModelForCausalLM.from_pretrained("Ransaka/sinhala-gpt2")
|
43 |
+
generator("දුර") #දුර ඈත පාසැල් වියේ පසුවූයේ මෙම සිද්ධිය සම්බන්ධයෙන් විමර්ශන සිදුකරන බවයි
|
|
|
44 |
```
|
45 |
or using git
|
46 |
```bash
|
47 |
git lfs install
|
48 |
+
git clone https://huggingface.co/Ransaka/sinhala-gpt2
|
49 |
```
|
50 |
|
51 |
### Training hyperparameters
|
|
|
73 |
- Transformers 4.26.1
|
74 |
- Pytorch 1.13.0
|
75 |
- Datasets 2.1.0
|
76 |
+
- Tokenizers 0.13.2
|