Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
widget:
|
3 |
+
- text: Em uma bela manhã de
|
4 |
+
- text: Em uma cidade tão grande como
|
5 |
+
- text: Maria e Joana são
|
6 |
+
license: mit
|
7 |
+
datasets:
|
8 |
+
- mc4
|
9 |
+
language:
|
10 |
+
- pt
|
11 |
+
metrics:
|
12 |
+
- perplexity
|
13 |
+
library_name: transformers
|
14 |
+
pipeline_tag: text-generation
|
15 |
+
---
|
16 |
+
# OPT-125M finetuned Portuguese
|
17 |
+
|
18 |
+
Fine-tuning the [OPT-125M](https://huggingface.co/facebook/opt-125m) model on a reduced corpus of MC4-Portuguese with approximately 300M tokens.
|
19 |
+
|
20 |
+
In this training a sequence length of 512 tokens was used, batch of 32 for 2 epochs.
|
21 |
+
|
22 |
+
With an A100 with 40GB of RAM, the training took around 3 hours
|
23 |
+
|
24 |
+
**Perplexity:** 9.4
|
25 |
+
|
26 |
+
## Sample Use
|
27 |
+
|
28 |
+
```python
|
29 |
+
from transformers import pipeline
|
30 |
+
generator = pipeline('text-generation', model='Mirelle/opt-125M-pt-br-finetuned', max_length=100, do_sample=True)
|
31 |
+
generator("Em uma bela manhã de")
|
32 |
+
```
|