Update README.md

README.md

@@ -10,66 +10,93 @@ base_model:
 - openai/whisper-small
 ---
 
-<!-- This model card has been generated automatically according to the information the Trainer had access to. You
-should probably proofread and complete it, then remove this comment. -->
+# **ScreenTalk**
+**Fine-tuned version of `openai/whisper-small` on the `DataLabX/ScreenTalk-XS` dataset**
+
+## **Model Summary**
+ScreenTalk is a fine-tuned version of OpenAI's Whisper-Small model, specifically trained for speech-to-text transcription using the **DataLabX/ScreenTalk-XS** dataset. The model is optimized to improve automatic speech recognition (ASR) performance in its target domain.
+
+On the evaluation set, it achieves:
+- **Loss**: `0.375`
+- **Word Error Rate (WER)**: `21.27%`
+
+## **Intended Uses & Limitations**
+### **Intended Use Cases**
+- **Speech-to-text transcription** for audio in the domain covered by `ScreenTalk-XS`
+- **Automatic subtitling** and **audio content analysis**
+- **Voice-assisted applications** where accurate ASR is needed
+
+### **Limitations**
+- May not generalize well to **out-of-domain** data
+- Performance is dependent on **audio quality** and **background noise**
+- The model is optimized for English (or the target language in `ScreenTalk-XS`)
+
+## **Training and Evaluation Data**
+The model was fine-tuned on the `DataLabX/ScreenTalk-XS` dataset, which contains domain-specific speech recordings. The dataset has been preprocessed and formatted to enhance ASR capabilities in specific contexts.
+
+## **Training Procedure**
+### **Hyperparameters**
+The model was trained with the following hyperparameters:
+
+| Hyperparameter                | Value        |
+|--------------------------------|-------------|
+| Learning Rate                  | `5e-05`     |
+| Train Batch Size               | `8`         |
+| Eval Batch Size                | `8`         |
+| Seed                            | `42`        |
+| Gradient Accumulation Steps     | `8`         |
+| Total Train Batch Size          | `64`        |
+| Optimizer                       | `AdamW` (β1=0.9, β2=0.999, ε=1e-08) |
+| Learning Rate Scheduler         | `Linear`    |
+| Warmup Steps                    | `10`        |
+| Total Training Steps            | `200`       |
+
+### **Training Progress**
+The model was trained for **200 steps**, and the WER improved over time:
+
+| Step | Training Loss | Validation Loss | WER (%) |
+|------|--------------|----------------|---------|
+| 20   | 1.1515       | 1.0011         | 22.33   |
+| 40   | 0.7024       | 0.6125         | 26.64   |
+| 60   | 0.3648       | 0.4175         | 23.00   |
+| 80   | 0.3753       | 0.3991         | 22.09   |
+| 100  | 0.3838       | 0.3952         | 22.83   |
+| 120  | 0.3358       | 0.3834         | 22.59   |
+| 140  | 0.1462       | 0.3924         | 22.01   |
+| 160  | 0.1636       | 0.3847         | 21.50   |
+| 180  | 0.1587       | 0.3778         | 21.36   |
+| 200  | 0.1583       | 0.3759         | 21.27   |
+
+## **Framework Versions**
+- **PEFT**: `0.14.0`
+- **Transformers**: `4.48.3`
+- **PyTorch**: `2.5.1+cu124`
+- **Datasets**: `3.3.2`
+- **Tokenizers**: `0.21.0`
+
+## **How to Use**
+To load and use this model for inference:
+
+```python
+from transformers import pipeline
+
+asr_pipeline = pipeline("automatic-speech-recognition", model="your_hf_username/ScreenTalk")
+audio_file = "path/to/audio.wav"
+
+transcription = asr_pipeline(audio_file)
+print(transcription["text"])
+```
+
+## **Citation**
+If you use this model, please cite:
+
+```java
+@misc{ScreenTalk,
+  title={ScreenTalk: A Fine-tuned Whisper-Small Model for Speech Recognition},
+  author={Your Name or Organization},
+  year={2025},
+  url={https://huggingface.co/your_hf_username/ScreenTalk}
+}
+```
 
-# ScreenTalk
 
-This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the DataLabX/ScreenTalk-XS dataset.
-It achieves the following results on the evaluation set:
-- Loss: 0.375
-- Wer: 21.27
-
-## Model description
-
-More information needed
-
-## Intended uses & limitations
-
-More information needed
-
-## Training and evaluation data
-
-More information needed
-
-## Training procedure
-
-### Training hyperparameters
-
-The following hyperparameters were used during training:
-- learning_rate: 5e-05
-- train_batch_size: 8
-- eval_batch_size: 8
-- seed: 42
-- gradient_accumulation_steps: 8
-- total_train_batch_size: 64
-- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
-- lr_scheduler_type: linear
-- lr_scheduler_warmup_steps: 10
-- training_steps: 200
-
-### Training results
-
-
-|Step           |	Training Loss|	Validation Loss|	Wer|
-|:-------------:|:-----:|:----:|:---------------:|
-|20	|1.151500	|1.001086	|22.332719|
-|40	|0.702400	|0.612483	|26.639884|
-|60	|0.364800	|0.417527	|23.004478|
-|80	|0.375300	|0.399105	|22.089041|
-|100	|0.383800	|0.395203	|22.833246|
-|120	|0.335800	|0.383432	|22.589568|
-|140	|0.146200	|0.392425	|22.010011|
-|160	|0.163600	|0.384719	|21.502898|
-|180	|0.158700	|0.377762	|21.364594|
-|200	|0.158300	|0.375860	|21.272392|
-
-
-
-### Framework versions
-
-- PEFT 0.14.0
-- Transformers 4.48.3
-- Pytorch 2.5.1+cu124
-- Datasets 3.3.2
-- Tokenizers 0.21.0