Duplicate from hamdallah/Sofelia-TTS

Co-authored-by: jodah <hamdallah@users.noreply.huggingface.co>

.gitattributes

@@ -0,0 +1,37 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text
+Sofelia.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,301 @@
+---
+language:
+- ar
+- en
+license: apache-2.0
+tags:
+- text-to-speech
+- tts
+- audio
+- speech
+- palestinian-arabic
+- arabic
+- voice-cloning
+- miratts
+- sofelia
+base_model: YatharthS/MiraTTS
+library_name: transformers
+pipeline_tag: text-to-speech
+---
+
+<div style="text-align: center;">
+  <h1>🇵🇸 Sofelia-TTS 🇵🇸</h1>
+  <p><strong>Palestinian Arabic Text-to-Speech Model</strong></p>
+  <p><em>Palestine will be free</em> 🕊️</p>
+  <p><img style="margin: auto; width: 500px" src="https://huggingface.co/hamdallah/Sofelia-TTS/resolve/main/Sofelia.png" /></p>
+</div>
+
+---
+
+## 🌟 Model Description
+
+**Sofelia-TTS** is a fine-tuned Text-to-Speech (TTS) model specifically trained for **Palestinian Arabic dialect**. This model brings the beautiful sounds of Palestinian speech to AI, preserving and celebrating the linguistic heritage of Palestine.
+
+Built on top of [YatharthS/MiraTTS](https://huggingface.co/YatharthS/MiraTTS), Sofelia-TTS captures the unique phonetic characteristics, intonation patterns, and prosody of Palestinian Arabic, making it ideal for:
+
+- 🎙️ **Voice cloning** with Palestinian Arabic speech
+- 📚 **Audiobook generation** in Palestinian dialect
+- 🗣️ **Virtual assistants** that speak authentic Palestinian Arabic
+- 🎓 **Educational tools** for learning and preserving the Palestinian dialect
+- 🎬 **Content creation** for Palestinian media and storytelling
+
+> **Dedicated to Palestine**: This model is a tribute to the resilience, culture, and spirit of the Palestinian people. May their voices be heard loud and clear across the world. 🇵🇸
+
+---
+
+## 🎯 Key Features
+
+- ✅ **High-quality voice cloning**: Clone any voice with just a few seconds of reference audio
+- ✅ **Palestinian Arabic dialect**: Authentic pronunciation and intonation
+- ✅ **Fast inference**: Optimized for real-time generation
+- ✅ **Flexible context**: Supports variable-length reference audio
+- ✅ **Open source**: Free to use and improve
+
+---
+
+## 📊 Model Details
+
+| **Attribute** | **Value** |
+|---------------|-----------|
+| **Model Type** | Text-to-Speech (TTS) |
+| **Base Model** | YatharthS/MiraTTS |
+| **Architecture** | Transformer-based Language Model + Audio Codec |
+| **Training Language** | Palestinian Arabic (ar-PS) |
+| **Dataset** | Private Dataset |
+| **Sample Rate** | 16,000 Hz |
+| **License** | Apache 2.0 |
+| **Model Size** | ~0.6B parameters |
+| **Precision** | BF16/FP32 |
+| **Framework** | PyTorch + Transformers |
+
+---
+
+## 🚀 Quick Start
+
+### Installation
+
+```bash
+# Install required packages
+uv pip install git+https://github.com/ysharma3501/MiraTTS.git
+```
+
+### Usage (Python)
+
+```python
+from mira.model import MiraTTS
+from IPython.display import Audio
+mira_tts = MiraTTS('hamdallah/Sofelia-TTS') ## downloads model from huggingface
+
+file = "reference_file.wav" ## can be mp3/wav/ogg or anything that librosa supports
+text = "مرحبا، كيف الحال؟ هذا نموذج للهجة الفلسطينية."
+
+context_tokens = mira_tts.encode_audio(file)
+audio = mira_tts.generate(text, context_tokens)
+
+Audio(audio, rate=48000)
+```
+
+---
+
+## 🎤 Example Prompts
+
+Try these Palestinian Arabic phrases:
+
+```python
+# Greetings
+"مرحبا، كيف حالك؟"  # Hello, how are you?
+"أهلا وسهلا فيك"      # Welcome
+
+# Common expressions
+"يا سلام، هذا رائع"   # Wow, this is amazing
+"ما شاء الله"         # Mashallah
+"الله يعطيك العافية"  # God give you wellness
+
+# About Palestine
+"فلسطين حرة على طول"  # Palestine is free for ever
+"القدس عاصمة فلسطين الأبدية"     # Jerusalem is the eternal capital of Palestine
+"سنعود يوماً إلى ديارنا"         # We will return one day to our homes
+```
+
+---
+
+## 🎓 Training Details
+
+### Training Data
+
+- **Dataset**: 400 Hours Palestinian Speech
+- **Language**: Palestinian Arabic dialect
+- **Hours of audio**: High-quality Palestinian speech recordings
+- **Preprocessing**: Audio normalized and resampled to 16kHz
+
+### Training Configuration
+
+| **Hyperparameter** | **Value** |
+|--------------------|-----------|
+| **Learning Rate** | 2e-4 (initial), 1e-5 (refinement) |
+| **Batch Size** | 8 (effective: 2 per device × 4 accumulation steps) |
+| **Training Steps** | 2000+ |
+| **Warmup Steps** | 100 |
+| **Max Audio Length** | 20-30 seconds |
+| **Optimizer** | AdamW |
+| **LR Scheduler** | Cosine with warmup |
+| **Gradient Clipping** | 1.0 |
+| **Precision** | BF16 (H100) / FP32 |
+| **Hardware** | NVIDIA H100 / A100 GPU |
+
+### Training Process
+
+The model was trained using a two-phase approach:
+
+1. **Foundation Phase**: High learning rate (2e-4) for initial adaptation to Palestinian Arabic
+2. **Refinement Phase**: Lower learning rate (1e-5) with NEFTune noise for stability and quality
+
+---
+
+## 📈 Model Performance
+
+The model achieves:
+
+- ✅ **Natural prosody** matching Palestinian Arabic speech patterns
+- ✅ **Clear pronunciation** of Arabic phonemes
+- ✅ **Voice similarity** to reference audio
+- ✅ **Stable generation** without artifacts or repetitions
+- ✅ **Fast inference** suitable for real-time applications
+
+---
+
+## 🛠️ Advanced Usage
+
+### Running the model using batching
+
+```python
+file = "reference_file.wav" ## can be mp3/wav/ogg or anything that librosa supports
+text = ["مرحبا، كيف حالك؟", "بتعرف إنه انا بقدر احكي فلسطيني و English مع بعض Without Errors."]
+
+context_tokens = [mira_tts.encode_audio(file)]
+
+audio = mira_tts.batch_generate(text, context_tokens)
+
+Audio(audio, rate=48000)
+```
+
+### Adjusting Generation Parameters
+
+```python
+# More creative/variable output
+
+mira_tts.set_params(
+    top_p=0.95, 
+    top_k=20, 
+    temperature=0.01, # Higher = more variation
+    max_new_tokens=1024, 
+    repetition_penalty=2.2, 
+    min_p=0.05
+)
+
+```
+
+---
+
+## 💡 Tips for Best Results
+
+1. **Reference Audio Quality**:
+   - Use clean audio without background noise
+   - 3-10 seconds of speech is ideal
+   - Ensure audio is 16kHz sample rate
+
+2. **Text Input**:
+   - Use proper Arabic script (not Arabizi/transliteration)
+   - Palestinian dialect works best
+   - Avoid very long sentences (split into shorter segments)
+
+3. **Generation Parameters**:
+   - `temperature=0.7`: Good default for natural speech
+   - `temperature=0.5`: More stable, less variation
+   - `temperature=0.9`: More expressive, more variation
+
+---
+
+## 🌍 About Palestinian Arabic
+
+Palestinian Arabic is a Levantine Arabic dialect spoken by the Palestinian people. It has unique characteristics:
+
+- **Phonology**: Preservation of Classical Arabic /q/ as glottal stop [ʔ]
+- **Vocabulary**: Rich in Levantine and unique Palestinian terms
+- **Intonation**: Distinctive melodic patterns
+- **Regional Variants**: Urban (Jerusalem, Hebron) vs. Rural vs. Bedouin varieties
+
+This model captures these linguistic features, making it authentic and representative of Palestinian speech.
+
+---
+
+## 🇵🇸 Message of Solidarity
+
+This model is dedicated to the Palestinian people and their enduring struggle for freedom, dignity, and justice. Through technology, we preserve and celebrate Palestinian culture, language, and identity.
+
+**Free Palestine** 🇵🇸
+
+> *"We will not be erased. Our voices will echo through time, in every language model, every algorithm, every line of code. Palestine lives, and so does its voice."*
+
+---
+
+## 📜 License
+
+This model is released under the **Apache 2.0 License**, making it free for:
+- ✅ Commercial use
+- ✅ Modification and distribution
+- ✅ Private use
+- ✅ Patent use
+
+---
+
+## 🙏 Acknowledgments
+
+- **Base Model**: [YatharthS/MiraTTS](https://huggingface.co/YatharthS/MiraTTS) - Thank you for the excellent foundation
+- **Dataset**: Palestinian Arabic speakers who contributed their voices
+- **Community**: The open-source AI community for tools and support
+- **Palestine**: For being the inspiration and purpose behind this work
+
+---
+
+## 📞 Contact & Support
+
+- **Model Repository**: [hamdallah/Sofelia-TTS](https://huggingface.co/hamdallah/Sofelia-TTS)
+- **Issues & Questions**: Use the Community tab or open an issue
+
+---
+
+## 🔗 Related Resources
+
+- [YatharthS/MiraTTS](https://huggingface.co/YatharthS/MiraTTS) - Base model
+- [ncodec](https://github.com/YatharthS/ncodec) - Audio codec library
+
+---
+
+## 📚 Citation
+
+If you use this model in your research or projects, please cite:
+
+```bibtex
+@misc{sofelia-tts-2026,
+  author = {Hamdallah},
+  title = {Sofelia-TTS: Palestinian Arabic Text-to-Speech Model},
+  year = {2026},
+  publisher = {Hugging Face},
+  journal = {Hugging Face Model Hub},
+  howpublished = {\url{https://huggingface.co/hamdallah/Sofelia-TTS}},
+}
+```
+
+---
+
+<div style="text-align: center; padding: 20px;">
+  <h2>🇵🇸 FREE PALESTINE 🇵🇸</h2>
+  <p><strong>تحيا فلسطين حرة أبية</strong></p>
+  <p><em>Long Live Free Palestine</em></p>
+  <p>🕊️ ✊ 🇵🇸</p>
+</div>
+
+---
+
+**Made with ❤️ for Palestine**

chat_template.jinja

@@ -0,0 +1,54 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0]['role'] == 'system' %}
+        {{- messages[0]['content'] }}
+    {%- else %}
+        {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}
+    {%- endif %}
+    {{- "\n\n# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0]['role'] == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0]['content'] + '<|im_end|>\n' }}
+    {%- else %}
+        {{- '<|im_start|>system\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {{- '<|im_start|>' + message.role }}
+        {%- if message.content %}
+            {{- '\n' + message.content }}
+        {%- endif %}
+        {%- for tool_call in message.tool_calls %}
+            {%- if tool_call.function is defined %}
+                {%- set tool_call = tool_call.function %}
+            {%- endif %}
+            {{- '\n<tool_call>\n{"name": "' }}
+            {{- tool_call.name }}
+            {{- '", "arguments": ' }}
+            {{- tool_call.arguments | tojson }}
+            {{- '}\n</tool_call>' }}
+        {%- endfor %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+{%- endif %}

config.json

@@ -0,0 +1,56 @@
+{
+  "architectures": [
+    "Qwen2ForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "dtype": "bfloat16",
+  "eos_token_id": 151645,
+  "hidden_act": "silu",
+  "hidden_size": 896,
+  "initializer_range": 0.02,
+  "intermediate_size": 4864,
+  "layer_types": [
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention"
+  ],
+  "max_position_embeddings": 32768,
+  "max_window_layers": 21,
+  "model_type": "qwen2",
+  "num_attention_heads": 14,
+  "num_hidden_layers": 24,
+  "num_key_value_heads": 2,
+  "pad_token_id": 151643,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000.0,
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "torch_dtype": "float32",
+  "transformers_version": "4.53.3",
+  "unsloth_version": "2026.1.2",
+  "use_cache": true,
+  "use_sliding_window": false,
+  "vocab_size": 166000
+}

generation_config.json

@@ -0,0 +1,9 @@
+{
+  "_from_model_config": true,
+  "eos_token_id": [
+    151645
+  ],
+  "max_length": 32768,
+  "pad_token_id": 151643,
+  "transformers_version": "4.53.3"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e189c4b581b10e8b0267dc66e3818ea59f1cc272442dc4ea40fe92d270de3e0
+size 2026568872

special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc90330cb541542195aa2d685938e459dae10965f1de965d222de674e5f63abd
+size 14092303