Text Classification
Transformers
Safetensors
Indonesian
bert
sentiment-analysis
indonesian
text-embeddings-inference
Instructions to use Bangkah/atha-text-classifier with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Bangkah/atha-text-classifier with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-classification", model="Bangkah/atha-text-classifier")# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForSequenceClassification tokenizer = AutoTokenizer.from_pretrained("Bangkah/atha-text-classifier") model = AutoModelForSequenceClassification.from_pretrained("Bangkah/atha-text-classifier", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -21,6 +21,12 @@ Model ini adalah fine-tuned indobenchmark/indobert-base-p1 untuk klasifikasi sen
|
|
| 21 |
|
| 22 |
Training data: https://huggingface.co/datasets/Bangkah/atha-text-dataset
|
| 23 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
## Validation Metrics
|
| 25 |
|
| 26 |
- Loss: 0.0004
|
|
@@ -49,3 +55,29 @@ Training data: https://huggingface.co/datasets/Bangkah/atha-text-dataset
|
|
| 49 |
weighted avg 1.0000 1.0000 1.0000 300
|
| 50 |
|
| 51 |
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 21 |
|
| 22 |
Training data: https://huggingface.co/datasets/Bangkah/atha-text-dataset
|
| 23 |
|
| 24 |
+
Label:
|
| 25 |
+
|
| 26 |
+
- 0: negative
|
| 27 |
+
- 1: neutral
|
| 28 |
+
- 2: positive
|
| 29 |
+
|
| 30 |
## Validation Metrics
|
| 31 |
|
| 32 |
- Loss: 0.0004
|
|
|
|
| 55 |
weighted avg 1.0000 1.0000 1.0000 300
|
| 56 |
|
| 57 |
```
|
| 58 |
+
|
| 59 |
+
## Intended Use
|
| 60 |
+
|
| 61 |
+
- Cocok untuk demo NLP end-to-end (dataset, training, deployment, API).
|
| 62 |
+
- Cocok untuk baseline eksperimen klasifikasi sentimen Bahasa Indonesia.
|
| 63 |
+
|
| 64 |
+
## Limitations
|
| 65 |
+
|
| 66 |
+
- Dataset saat ini masih sintetis, sehingga metrik tinggi belum mencerminkan performa dunia nyata.
|
| 67 |
+
- Model bisa sensitif pada pola kalimat template dari generator.
|
| 68 |
+
- Untuk produksi, wajib evaluasi ulang dengan data real dan data out-of-domain.
|
| 69 |
+
|
| 70 |
+
## Training Setup
|
| 71 |
+
|
| 72 |
+
- Base model: indobenchmark/indobert-base-p1
|
| 73 |
+
- Epoch: 1
|
| 74 |
+
- Train split: 1500
|
| 75 |
+
- Validation split: 300
|
| 76 |
+
- Metric utama: Accuracy dan Macro F1
|
| 77 |
+
|
| 78 |
+
## Error Analysis
|
| 79 |
+
|
| 80 |
+
Laporan detail kesalahan tersedia di:
|
| 81 |
+
|
| 82 |
+
- `error_analysis.md`
|
| 83 |
+
- `error_analysis.csv`
|