File size: 4,000 Bytes
a0cca4a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 |
Here's your model card based on the example you provided:
---
# RoBERTa-Base for News Classification (FP16 Quantized)
This is a RoBERTa-Base model fine-tuned on the **AG News dataset** for text classification. It categorizes news articles into one of four classes: **World, Sports, Business, and Science/Technology**. The model has been further **quantized to FP16** for improved inference speed and reduced memory usage, making it efficient for deployment on resource-constrained environments.
---
## **Model Details**
### **Model Description**
- **Model Type:** Transformer-based text classifier
- **Base Model:** `roberta-base`
- **Fine-Tuned Dataset:** AG News
- **Maximum Sequence Length:** 512 tokens
- **Output:** One of four news categories
- **Task:** Text classification
---
## **Full Model Architecture**
```python
RobertaForSequenceClassification(
(roberta): RobertaModel(
(embeddings): RobertaEmbeddings(...)
(encoder): RobertaEncoder(...)
)
(classifier): RobertaClassificationHead(
(dense): Linear(in_features=768, out_features=768, bias=True)
(dropout): Dropout(p=0.1)
(out_proj): Linear(in_features=768, out_features=4, bias=True)
)
)
```
---
## **Usage Instructions**
### **Installation**
```bash
pip install -U transformers torch
```
### **Loading the Model for Inference**
```python
from transformers import RobertaForSequenceClassification, RobertaTokenizer
import torch
# Load the model and tokenizer
model_name = "AventIQ-AI/Roberta-Base-News-Classification" # Update with your model ID
tokenizer = RobertaTokenizer.from_pretrained(model_name)
model = RobertaForSequenceClassification.from_pretrained(model_name)
# Move to GPU if available
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
# Function to predict category
def predict(text):
inputs = tokenizer(text, return_tensors="pt", padding="max_length", truncation=True, max_length=512).to(device)
with torch.no_grad():
outputs = model(**inputs)
logits = outputs.logits
predicted_class = torch.argmax(logits, dim=1).item()
class_labels = {0: "World", 1: "Sports", 2: "Business", 3: "Science/Technology"}
return class_labels[predicted_class]
# Example usage
custom_text = "Stock prices are rising due to global economic recovery."
predicted_label = predict(custom_text)
print(f"Predicted Category: {predicted_label}")
```
---
## **Training Details**
### **Training Dataset**
- **Name:** AG News
- **Size:** 50,000 training samples, 7,600 test samples
- **Labels:**
- **0:** World
- **1:** Sports
- **2:** Business
- **3:** Science/Technology
---
## **Training Hyperparameters**
### **Non-Default Hyperparameters:**
- **per_device_train_batch_size:** 8
- **per_device_eval_batch_size:** 8
- **gradient_accumulation_steps:** 2 (effective batch size = 16)
- **num_train_epochs:** 3
- **learning_rate:** 2e-5
- **fp16:** True (for reduced memory footprint and faster inference)
- **weight_decay:** 0.01
- **optimizer:** AdamW
---
## **Model Performance**
| Metric | Score |
|---------|-------|
| Accuracy | **94.3%** |
| F1 Score | **94.1%** |
| Precision | **94.5%** |
| Recall | **94.2%** |
*(Update these values based on your actual evaluation results.)*
---
## **Quantization Details**
- The model has been quantized to **FP16** to reduce its size and improve inference speed.
- FP16 quantization provides a **2x reduction in memory** while maintaining similar accuracy.
---
## **Limitations & Considerations**
- The model is **trained on AG News** and may not generalize well to **other domains** such as medical, legal, or entertainment news.
- Due to **FP16 quantization**, there might be a minor loss in precision, but inference speed is significantly improved.
- The model is **not intended for real-time misinformation detection**—it only classifies text based on its most probable category.
--- |