Update README.md

README.md

@@ -1,147 +1,66 @@
-Here’s the improved **README.md** with proper YAML metadata and a table for epochs, loss, and validation loss.
-
----
-
-```yaml
 ---
 library_name: transformers
 language:
-  - ur
+- ur
 license: apache-2.0
 base_model: openai/whisper-medium
 tags:
-  - automatic-speech-recognition
-  - urdu
-  - whisper
-  - fine-tuned
+- generated_from_trainer
 datasets:
-  - fsicoli/common_voice_19_0
+- fsicoli/common_voice_19_0
 model-index:
-  - name: whisper-medium-ur-v2
-    results:
-      - task:
-          type: automatic-speech-recognition
-        dataset:
-          name: Common Voice 19.0
-          type: fsicoli/common_voice_19_0
-        metrics:
-          - name: Validation Loss
-            type: loss
-            value: 0.3571
-          - name: Word Error Rate (WER)
-            type: wer
-            value: 25.17
+- name: Whisper Medium Ur - Your Name
+  results: []
 ---
-```
 
-# 🚀 Whisper Medium Urdu (whisper-medium-ur-v2)
+<!-- 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. -->
 
-Fine-tuned version of [openai/whisper-medium](https://huggingface.co/openai/whisper-medium) for **Urdu Automatic Speech Recognition (ASR)** on the **Common Voice 19.0** dataset.
+# Whisper Medium Ur - Your Name
 
----
+This model is a fine-tuned version of [openai/whisper-medium](https://huggingface.co/openai/whisper-medium) on the Common Voice 19.0 dataset.
+It achieves the following results on the evaluation set:
+- eval_loss: 0.3571
+- eval_wer: 25.1658
+- eval_runtime: 4297.3715
+- eval_samples_per_second: 1.167
+- eval_steps_per_second: 0.146
+- epoch: 1.3108
+- step: 1000
 
-## 📊 Training Performance
-
-| Epoch | Training Loss | Validation Loss | WER (%) |
-|-------|--------------|----------------|---------|
-| 0.5   | 0.4503       | 0.4121         | 28.45   |
-| 1.0   | 0.2304       | 0.3582         | 25.29   |
-| 1.31  | 0.1733       | 0.3571         | 25.17   |
-
----
+## Model description
 
-## 📌 Model Description  
-This model is based on **Whisper Medium**, a transformer-based sequence-to-sequence ASR model trained by OpenAI. It has been **fine-tuned on Urdu speech data** to improve transcription accuracy.
+More information needed
 
-- **Base Model:** [openai/whisper-medium](https://huggingface.co/openai/whisper-medium)  
-- **Language:** Urdu (ur)  
-- **Dataset:** [Common Voice 19.0](https://huggingface.co/datasets/fsicoli/common_voice_19_0)  
-
----
-
-## 🚀 Intended Use & Limitations  
-✅ **Best suited for:**  
-- Urdu speech-to-text transcription  
-- Conversational & broadcast speech  
-
-⚠️ **Limitations:**  
-- May struggle with **noisy environments**  
-- Accuracy depends on **audio quality and speaker accents**  
-- Not tested for **code-switching (mixing Urdu with English)**  
-
----
+## Intended uses & limitations
 
-## 🔧 How to Use  
-You can use this model with 🤗 Transformers:
+More information needed
 
-```python
-from transformers import pipeline
+## Training and evaluation data
 
-pipe = pipeline("automatic-speech-recognition", model="GogetaBlueMUI/whisper-medium-ur-v2")
-result = pipe("path_to_audio.wav")
-print(result["text"])
-```
+More information needed
 
-For lower-level inference:
+## Training procedure
 
-```python
-from transformers import WhisperProcessor, WhisperForConditionalGeneration
-import torch
-import torchaudio
+### Training hyperparameters
 
-# Load model & processor
-processor = WhisperProcessor.from_pretrained("GogetaBlueMUI/whisper-medium-ur-v2")
-model = WhisperForConditionalGeneration.from_pretrained("GogetaBlueMUI/whisper-medium-ur-v2")
+The following hyperparameters were used during training:
+- learning_rate: 3e-06
+- train_batch_size: 8
+- eval_batch_size: 8
+- seed: 42
+- gradient_accumulation_steps: 2
+- total_train_batch_size: 16
+- optimizer: Use 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: 100
+- training_steps: 1000
+- mixed_precision_training: Native AMP
 
-# Load audio file
-waveform, sample_rate = torchaudio.load("path_to_audio.wav")
-inputs = processor(waveform, sampling_rate=sample_rate, return_tensors="pt")
-
-# Generate transcription
-with torch.no_grad():
-    predicted_ids = model.generate(inputs.input_features)
-transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)[0]
-
-print(transcription)
-```
-
----
-
-## 🛠 Training Details  
-### **Hyperparameters**  
-- **Learning Rate:** 3e-6  
-- **Batch Size:** 8 (per device)  
-- **Gradient Accumulation Steps:** 2  
-- **Optimizer:** AdamW  
-- **Scheduler:** Linear Warmup (100 steps)  
-- **Precision:** Mixed Precision (AMP)  
-
-### **Framework Versions**  
-- **Transformers:** 4.49.0  
-- **PyTorch:** 2.5.1+cu121  
-- **Datasets:** 3.4.1  
-- **Tokenizers:** 0.21.0  
-
----
-
-## 📚 Citation  
-If you use this model in your research or project, please cite:
-
-```bibtex
-@misc{whisper-medium-ur-v2,
-  author = {GogetaBlueMUI},
-  title = {Whisper Medium Urdu Fine-Tuned Model},
-  year = {2025},
-  publisher = {Hugging Face},
-  url = {https://huggingface.co/GogetaBlueMUI/whisper-medium-ur-v2}
-}
-```
-
----
+### Framework versions
 
-## ⭐ Acknowledgements  
-- **OpenAI** for the Whisper model  
-- **Mozilla Common Voice** for the dataset  
-- **Hugging Face** for hosting  
+- Transformers 4.49.0
+- Pytorch 2.5.1+cu121
+- Datasets 3.4.1
+- Tokenizers 0.21.0
 
----