End of training

README.md

@@ -1,78 +1,85 @@
 ---
 library_name: transformers
+language:
+- ur
 license: apache-2.0
-base_model: GogetaBlueMUI/whisper-medium-ur-jalandhary
+base_model: openai/whisper-medium
 tags:
-- automatic-speech-recognition
-- ASR
-- Urdu
-- Whisper
-- speech-to-text
 - generated_from_trainer
 datasets:
-- common_voice_11_0
+- fsicoli/common_voice_19_0
 metrics:
 - wer
-inference: true
-widget:
-  - example_title: "Test Urdu Audio"
-    src: "https://huggingface.co/datasets/Narsil/asr_dummy/resolve/main/test.flac"
 model-index:
-- name: whisper-medium-ur
+- name: Whisper Medium Ur - Your Name
   results:
   - task:
       name: Automatic Speech Recognition
       type: automatic-speech-recognition
     dataset:
-      name: common_voice_11_0
-      type: common_voice_11_0
+      name: Common Voice 19.0
+      type: fsicoli/common_voice_19_0
       config: ur
       split: test
-      args: ur
+      args: 'config: ur, split: test'
     metrics:
     - name: Wer
       type: wer
-      value: 0.2744
+      value: 27.349454082657914
 ---
 
-# 🗣️ Whisper-Medium-Ur: Urdu Speech Recognition Model
+<!-- 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. -->
 
-This model is a fine-tuned version of **[GogetaBlueMUI/whisper-medium-ur-jalandhary](https://huggingface.co/GogetaBlueMUI/whisper-medium-ur-jalandhary)** on the **Common Voice 11.0 Urdu dataset**. It is designed for **automatic speech recognition (ASR)** in Urdu and achieves the following results on the evaluation set:
+# Whisper Medium Ur - Your Name
 
-- **Loss:** 0.5375  
-- **WER (Word Error Rate):** 27.44%  
-- **CER (Character Error Rate):** 12.37%  
+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:
+- Loss: 0.3613
+- Wer: 27.3495
 
----
+## Model description
 
-## **📌 Model Description**
-- The model is based on **OpenAI's Whisper-Medium**.
-- It is fine-tuned specifically for **Urdu speech transcription**.
-- Works best on **clear audio recordings** with minimal background noise.
+More information needed
 
----
+## Intended uses & limitations
 
-## **🛠️ Intended Uses & Limitations**
-### ✅ **Intended Uses**
-- **Transcribing Urdu speech** into text.
-- **Generating subtitles** for Urdu videos.
-- **Building Urdu speech-to-text applications**.
+More information needed
 
-### ❌ **Limitations**
-- May struggle with **noisy environments**.
-- May not perform well on **regional Urdu dialects**.
-- Limited **code-mixing** support (Urdu + English).
+## Training and evaluation data
 
----
+More information needed
+
+## Training procedure
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 5e-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: 40
+- training_steps: 800
+- mixed_precision_training: Native AMP
+
+### Training results
 
-## **💻 Usage**
-You can use this model with the **Hugging Face Transformers pipeline**:
+| Training Loss | Epoch  | Step | Validation Loss | Wer     |
+|:-------------:|:------:|:----:|:---------------:|:-------:|
+| 0.5093        | 0.2623 | 200  | 0.4290          | 29.3009 |
+| 0.4283        | 0.5246 | 400  | 0.3918          | 29.4996 |
+| 0.4435        | 0.7869 | 600  | 0.3705          | 27.1239 |
+| 0.2939        | 1.0485 | 800  | 0.3613          | 27.3495 |
 
-```python
-from transformers import pipeline
 
-pipe = pipeline("automatic-speech-recognition", model="GogetaBlueMUI/whisper-medium-ur")
+### Framework versions
 
-# Run inference on an audio file
-result = pipe("path/to/your_audio_file.wav")
-print(result["text"])
+- Transformers 4.49.0
+- Pytorch 2.5.1+cu121
+- Datasets 3.4.0
+- Tokenizers 0.21.0

generation_config.json

@@ -246,5 +246,5 @@
     "transcribe": 50359,
     "translate": 50358
   },
-  "transformers_version": "4.47.0"
+  "transformers_version": "4.49.0"
 }