fawzanaramam
commited on
Update README.md
Browse files
README.md
CHANGED
@@ -4,7 +4,11 @@ language:
|
|
4 |
license: apache-2.0
|
5 |
base_model: openai/whisper-small
|
6 |
tags:
|
7 |
-
-
|
|
|
|
|
|
|
|
|
8 |
datasets:
|
9 |
- fawzanaramam/the-truth-1st-chapter
|
10 |
metrics:
|
@@ -20,67 +24,76 @@ model-index:
|
|
20 |
type: fawzanaramam/the-truth-1st-chapter
|
21 |
args: 'config: ar, split: train'
|
22 |
metrics:
|
23 |
-
- name:
|
24 |
type: wer
|
25 |
value: 0.0
|
26 |
---
|
27 |
|
28 |
-
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
29 |
-
should probably proofread and complete it, then remove this comment. -->
|
30 |
-
|
31 |
# Whisper Small Finetuned on Surah Fatiha
|
32 |
|
33 |
-
This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small)
|
34 |
-
|
35 |
-
|
36 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
37 |
|
38 |
-
|
39 |
|
40 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
41 |
|
42 |
-
## Intended
|
43 |
|
44 |
-
|
|
|
|
|
|
|
45 |
|
46 |
-
|
|
|
|
|
|
|
47 |
|
48 |
-
|
49 |
|
50 |
-
|
51 |
|
52 |
-
|
53 |
|
|
|
54 |
The following hyperparameters were used during training:
|
55 |
-
-
|
56 |
-
-
|
57 |
-
-
|
58 |
-
-
|
59 |
-
-
|
60 |
-
-
|
61 |
-
-
|
62 |
-
-
|
63 |
-
-
|
64 |
-
|
65 |
-
###
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
| No log | 1.1111 | 20 | 0.3582 | 29.7872 |
|
71 |
-
| 0.6771 | 1.6667 | 30 | 0.1882 | 23.4043 |
|
72 |
-
| 0.6771 | 2.2222 | 40 | 0.0928 | 25.0 |
|
73 |
-
| 0.0289 | 2.7778 | 50 | 0.0660 | 34.0426 |
|
74 |
-
| 0.0289 | 3.3333 | 60 | 0.0484 | 32.9787 |
|
75 |
-
| 0.0289 | 3.8889 | 70 | 0.0241 | 25.5319 |
|
76 |
-
| 0.0056 | 4.4444 | 80 | 0.0184 | 28.7234 |
|
77 |
-
| 0.0056 | 5.0 | 90 | 0.0111 | 0.0 |
|
78 |
-
| 0.0019 | 5.5556 | 100 | 0.0088 | 0.0 |
|
79 |
-
|
80 |
-
|
81 |
-
### Framework versions
|
82 |
-
|
83 |
-
- Transformers 4.41.1
|
84 |
-
- Pytorch 2.2.1+cu121
|
85 |
-
- Datasets 2.19.1
|
86 |
-
- Tokenizers 0.19.1
|
|
|
4 |
license: apache-2.0
|
5 |
base_model: openai/whisper-small
|
6 |
tags:
|
7 |
+
- fine-tuned
|
8 |
+
- Quran
|
9 |
+
- automatic-speech-recognition
|
10 |
+
- arabic
|
11 |
+
- whisper
|
12 |
datasets:
|
13 |
- fawzanaramam/the-truth-1st-chapter
|
14 |
metrics:
|
|
|
24 |
type: fawzanaramam/the-truth-1st-chapter
|
25 |
args: 'config: ar, split: train'
|
26 |
metrics:
|
27 |
+
- name: Word Error Rate (WER)
|
28 |
type: wer
|
29 |
value: 0.0
|
30 |
---
|
31 |
|
|
|
|
|
|
|
32 |
# Whisper Small Finetuned on Surah Fatiha
|
33 |
|
34 |
+
This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small), transcribing Surah Fatiha, the first chapter of the Quran. It has been trained using *The Truth 2.0 - Surah Fatiha* dataset and achieves excellent results with a Word Error Rate (WER) of **0.0**, indicating perfect transcription on the evaluation set.
|
35 |
+
|
36 |
+
## Model Description
|
37 |
+
|
38 |
+
Whisper Small is a transformer-based automatic speech recognition (ASR) model developed by OpenAI. By fine-tuning it on the *Surah Fatiha* dataset, this model becomes highly accurate in transcribing Quranic recitation. It is designed to assist in religious, educational, and research-oriented tasks that require precise Quranic transcription.
|
39 |
+
|
40 |
+
## Performance Metrics
|
41 |
+
|
42 |
+
On the evaluation set, the model achieved:
|
43 |
+
- **Loss**: 0.0088
|
44 |
+
- **Word Error Rate (WER)**: 0.0
|
45 |
+
|
46 |
+
These metrics showcase the model's exceptional performance and reliability in transcribing Surah Fatiha audio.
|
47 |
+
|
48 |
+
## Training Results
|
49 |
|
50 |
+
The following table summarizes the training process and results:
|
51 |
|
52 |
+
| **Training Loss** | **Epoch** | **Step** | **Validation Loss** | **WER** |
|
53 |
+
|:------------------:|:---------:|:--------:|:-------------------:|:----------:|
|
54 |
+
| No log | 0.5556 | 10 | 1.1057 | 96.2766 |
|
55 |
+
| No log | 1.1111 | 20 | 0.3582 | 29.7872 |
|
56 |
+
| 0.6771 | 1.6667 | 30 | 0.1882 | 23.4043 |
|
57 |
+
| 0.6771 | 2.2222 | 40 | 0.0928 | 25.0 |
|
58 |
+
| 0.0289 | 2.7778 | 50 | 0.0660 | 34.0426 |
|
59 |
+
| 0.0289 | 3.3333 | 60 | 0.0484 | 32.9787 |
|
60 |
+
| 0.0289 | 3.8889 | 70 | 0.0241 | 25.5319 |
|
61 |
+
| 0.0056 | 4.4444 | 80 | 0.0184 | 28.7234 |
|
62 |
+
| 0.0056 | 5.0 | 90 | 0.0111 | 0.0 |
|
63 |
+
| 0.0019 | 5.5556 | 100 | 0.0088 | 0.0 |
|
64 |
|
65 |
+
## Intended Uses & Limitations
|
66 |
|
67 |
+
### Intended Uses
|
68 |
+
- **Speech-to-text transcription** of Quranic recitation for Surah Fatiha.
|
69 |
+
- Educational tools to assist in learning and practicing Quranic recitation.
|
70 |
+
- Research and analysis of Quranic audio transcription methods.
|
71 |
|
72 |
+
### Limitations
|
73 |
+
- This model is fine-tuned specifically for Surah Fatiha and may not generalize well to other chapters or non-Quranic Arabic audio.
|
74 |
+
- Variability in audio quality, accents, or recitation styles might affect performance.
|
75 |
+
- Optimal performance is achieved with high-quality audio inputs.
|
76 |
|
77 |
+
## Training and Evaluation Data
|
78 |
|
79 |
+
The model was trained on *The Truth 2.0 - Surah Fatiha* dataset, which comprises high-quality audio recordings of Surah Fatiha and their corresponding transcripts. The dataset was meticulously curated to ensure the accuracy and authenticity of Quranic content.
|
80 |
|
81 |
+
## Training Procedure
|
82 |
|
83 |
+
### Training Hyperparameters
|
84 |
The following hyperparameters were used during training:
|
85 |
+
- **Learning Rate**: 1e-05
|
86 |
+
- **Training Batch Size**: 16
|
87 |
+
- **Evaluation Batch Size**: 8
|
88 |
+
- **Seed**: 42
|
89 |
+
- **Optimizer**: Adam (betas=(0.9, 0.999), epsilon=1e-08)
|
90 |
+
- **Learning Rate Scheduler**: Linear
|
91 |
+
- **Warmup Steps**: 10
|
92 |
+
- **Training Steps**: 100
|
93 |
+
- **Mixed Precision Training**: Native AMP
|
94 |
+
|
95 |
+
### Framework Versions
|
96 |
+
- **Transformers**: 4.41.1
|
97 |
+
- **PyTorch**: 2.2.1+cu121
|
98 |
+
- **Datasets**: 2.19.1
|
99 |
+
- **Tokenizers**: 0.19.1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|