Update README.md
Browse files
README.md
CHANGED
|
@@ -6,34 +6,64 @@ tags:
|
|
| 6 |
- generated_from_trainer
|
| 7 |
metrics:
|
| 8 |
- accuracy
|
| 9 |
-
|
| 10 |
-
-
|
| 11 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
---
|
| 13 |
|
| 14 |
-
|
| 15 |
-
should probably proofread and complete it, then remove this comment. -->
|
| 16 |
|
| 17 |
-
|
|
|
|
| 18 |
|
| 19 |
-
This model is a fine-tuned version of [agentlans/multilingual-e5-small-aligned-v2](https://huggingface.co/agentlans/multilingual-e5-small-aligned-v2)
|
| 20 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 21 |
- Loss: 0.1983
|
| 22 |
- Accuracy: 0.9515
|
| 23 |
- Combined Score: 1.3494
|
| 24 |
- Num Input Tokens Seen: 122880000
|
| 25 |
|
| 26 |
-
## Model description
|
| 27 |
-
|
| 28 |
-
More information needed
|
| 29 |
|
| 30 |
-
##
|
| 31 |
|
| 32 |
-
|
|
|
|
| 33 |
|
| 34 |
-
|
|
|
|
|
|
|
| 35 |
|
| 36 |
-
|
|
|
|
|
|
|
|
|
|
| 37 |
|
| 38 |
## Training procedure
|
| 39 |
|
|
@@ -62,4 +92,4 @@ The following hyperparameters were used during training:
|
|
| 62 |
- Transformers 4.51.3
|
| 63 |
- Pytorch 2.6.0+cu124
|
| 64 |
- Datasets 3.2.0
|
| 65 |
-
- Tokenizers 0.21.0
|
|
|
|
| 6 |
- generated_from_trainer
|
| 7 |
metrics:
|
| 8 |
- accuracy
|
| 9 |
+
language:
|
| 10 |
+
- ar
|
| 11 |
+
- zh
|
| 12 |
+
- cs
|
| 13 |
+
- da
|
| 14 |
+
- nl
|
| 15 |
+
- fr
|
| 16 |
+
- de
|
| 17 |
+
- el
|
| 18 |
+
- hu
|
| 19 |
+
- id
|
| 20 |
+
- it
|
| 21 |
+
- ja
|
| 22 |
+
- fa
|
| 23 |
+
- pl
|
| 24 |
+
- pt
|
| 25 |
+
- ru
|
| 26 |
+
- es
|
| 27 |
+
- sv
|
| 28 |
+
- tr
|
| 29 |
+
- vi
|
| 30 |
+
datasets:
|
| 31 |
+
- agentlans/fineweb2hq-vs-c4
|
| 32 |
+
pipeline_tag: text-classification
|
| 33 |
---
|
| 34 |
|
| 35 |
+
# agentlans/multilingual-e5-small-fineweb2hq-vs-c4-classifier
|
|
|
|
| 36 |
|
| 37 |
+
> [!IMPORTANT]
|
| 38 |
+
> **Note:** This model is provided for reference and reproducibility, not for standalone use.
|
| 39 |
|
| 40 |
+
This model is a fine-tuned version of [agentlans/multilingual-e5-small-aligned-v2](https://huggingface.co/agentlans/multilingual-e5-small-aligned-v2)
|
| 41 |
+
on the [agentlans/fineweb2hq-vs-c4](https://huggingface.co/datasets/agentlans/fineweb2hq-vs-c4) dataset.
|
| 42 |
+
|
| 43 |
+
The aim is to classify text as higher quality (FineWeb 2 HQ) or lower quality (C4) for AI training.
|
| 44 |
+
|
| 45 |
+
Training dataset:
|
| 46 |
+
|
| 47 |
+
On the validation set:
|
| 48 |
- Loss: 0.1983
|
| 49 |
- Accuracy: 0.9515
|
| 50 |
- Combined Score: 1.3494
|
| 51 |
- Num Input Tokens Seen: 122880000
|
| 52 |
|
|
|
|
|
|
|
|
|
|
| 53 |
|
| 54 |
+
## Example
|
| 55 |
|
| 56 |
+
```python
|
| 57 |
+
from transformers import pipeline
|
| 58 |
|
| 59 |
+
classifier = pipeline("text-classification", model="agentlans/multilingual-e5-small-fineweb2hq-vs-c4-classifier")
|
| 60 |
+
classifier("Your text here.")
|
| 61 |
+
```
|
| 62 |
|
| 63 |
+
## Limitations
|
| 64 |
+
- **Not trained on English data**
|
| 65 |
+
- Tends to be overly permissive, labelling most texts outside training data as high quality
|
| 66 |
+
- May be biased against some text types
|
| 67 |
|
| 68 |
## Training procedure
|
| 69 |
|
|
|
|
| 92 |
- Transformers 4.51.3
|
| 93 |
- Pytorch 2.6.0+cu124
|
| 94 |
- Datasets 3.2.0
|
| 95 |
+
- Tokenizers 0.21.0
|