71 lines
1.4 KiB
Markdown
71 lines
1.4 KiB
Markdown
|
---
|
||
|
|
language:
|
||
|
|
- pt
|
||
|
|
metrics:
|
||
|
|
- perplexity
|
||
|
|
pipeline_tag: text-generation
|
||
|
|
---
|
||
|
|
# Model Card for Model ID
|
||
|
|
|
||
|
|
A Portuguese language model trained on https://huggingface.co/facebook/opt-125m .
|
||
|
|
|
||
|
|
## Model Details
|
||
|
|
|
||
|
|
### Model Description
|
||
|
|
|
||
|
|
<!-- Provide a longer summary of what this model is. -->
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
- **Developed by:** Monique Monteiro
|
||
|
|
- **Shared by [optional]:** Monique Monteiro
|
||
|
|
- **Model type:** OPT
|
||
|
|
- **Language(s) (NLP):** Portuguese
|
||
|
|
- **License:** [More Information Needed]
|
||
|
|
- **Finetuned from model [optional]:** facebook/opt-125m
|
||
|
|
|
||
|
|
Use the code below to get started with the model.
|
||
|
|
|
||
|
|
```python
|
||
|
|
generator = pipeline('text-generation', 'monilouise/opt125M_portuguese')
|
||
|
|
output = generator("Era uma vez", max_length=50, do_sample=True)
|
||
|
|
```
|
||
|
|
|
||
|
|
## Training Details
|
||
|
|
|
||
|
|
### Training Data
|
||
|
|
|
||
|
|
The model was trained on gs://unicamp-dl/ia025a_2022s1/aula9/sample-1gb.txt
|
||
|
|
|
||
|
|
### Training Procedure
|
||
|
|
|
||
|
|
The model was trained for 3 epochs, by using learning rate = 5e-5 (linear scheduler).
|
||
|
|
|
||
|
|
#### Preprocessing [optional]
|
||
|
|
|
||
|
|
All text was tokenized and broken into chunks of 1024 tokens.
|
||
|
|
|
||
|
|
|
||
|
|
#### Training Hyperparameters
|
||
|
|
|
||
|
|
- **Training regime:** fp16 mixed precision
|
||
|
|
|
||
|
|
#### Speeds, Sizes, Times [optional]
|
||
|
|
|
||
|
|
Training time: 17 hours
|
||
|
|
|
||
|
|
## Evaluation
|
||
|
|
|
||
|
|
The model was evaluated on a 5% validation split.
|
||
|
|
|
||
|
|
#### Metrics
|
||
|
|
|
||
|
|
Perplexity = 7.94.
|
||
|
|
|
||
|
|
## Model Card Authors [optional]
|
||
|
|
|
||
|
|
moniquelouise@gmail.com
|
||
|
|
|
||
|
|
## Model Card Contact
|
||
|
|
|
||
|
|
moniquelouise@gmail.com
|