File size: 1,880 Bytes
0e61314
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
language:
- en
license: mit
size_categories:
- 100K<n<1M
task_categories:
- text-classification
tags:
- climate
- environment
- fasttext
- fineweb-edu
- filtered
dataset_info:
  features:
  - name: text
    dtype: string
  - name: id
    dtype: string
  - name: url
    dtype: string
  - name: climate_prob
    dtype: float64
  - name: source
    dtype: string
pretty_name: FineWeb-Edu v2 Climate FastText Filtered
---

# FineWeb-Edu v2 - FastText Climate Filtered

A climate and environment-focused subset of [sraj/finewebedu-climate-v2](https://huggingface.co/datasets/sraj/finewebedu-climate-v2), further filtered using a trained **FastText binary classifier**.

## Overview

This dataset applies a supervised FastText climate classifier to the FineWeb-Edu climate v2 dataset. Each record includes a climate probability score from the classifier, providing a confidence measure for climate relevance.

## Pipeline

1. **Source**: [sraj/finewebedu-climate-v2](https://huggingface.co/datasets/sraj/finewebedu-climate-v2) (pre-filtered FineWeb-Edu)
2. **Classifier**: FastText supervised model trained on 10K GPT-labeled samples
3. **Threshold**: Records with climate_prob >= 0.5 are included

## Fields

| Field | Type | Description |
|-------|------|-------------|
| text | string | The document text |
| id | string | Original document ID |
| url | string | Source URL |
| climate_prob | float | FastText classifier probability (0-1) |
| source | string | Source dataset identifier |

## Usage

```python
from datasets import load_dataset

dataset = load_dataset("Michaelyya/fineweb-edu-v2-FastT")
```

## Model Details

- **Architecture**: FastText supervised classifier
- **Training**: 10,000 samples labeled via GPT weak supervision
- **Labels**: Binary classification (climate vs. other)
- **Hyperparameters**: lr=0.5, epochs=25, wordNgrams=2, dim=100

## License

MIT