RohanMuralidharan
/

Transync

@@ -1,38 +1,15 @@
 ---
-language:
-  - asm
-  - ben
-  - brx
-  - doi
-  - gom
-  - guj
-  - hin
-  - kan
-  - kas
-  - mai
-  - mal
-  - mar
-  - mni
-  - npi
-  - ory
-  - pan
-  - san
-  - sat
-  - snd
-  - tam
-  - tel
-  - urd
-  - eng
 license: mit
 library_name: transformers
-pipeline_tag: text2text-generation
 tags:
-  - translation
-  - multilingual
-  - indic
-  - mbart
-  - seq2seq
-  - huggingface
 ---
 # Transync
@@ -59,73 +36,33 @@ Transync is a multilingual translation model based on the MBart architecture, de
 The model supports 50+ languages including:
-### Indian Languages
-- Assamese (asm)
-- Bengali (ben)
-- Bodo (brx)
-- Dogri (doi)
-- Konkani (gom)
-- Gujarati (guj)
-- Hindi (hin)
-- Kannada (kan)
-- Kashmiri (kas)
-- Maithili (mai)
-- Malayalam (mal)
-- Marathi (mar)
-- Manipuri (mni)
-- Nepali (npi)
-- Odia (ory)
-- Punjabi (pan)
-- Sanskrit (san)
-- Santali (sat)
-- Sindhi (snd)
-- Tamil (tam)
-- Telugu (tel)
-- Urdu (urd)
-### Other Languages
-- Arabic (ara)
-- Czech (ces)
-- German (deu)
-- Spanish (spa)
-- Estonian (est)
-- Finnish (fin)
-- French (fra)
-- Hebrew (heb)
-- Croatian (hrv)
-- Indonesian (ind)
-- Italian (ita)
-- Japanese (jpn)
-- Georgian (kat)
-- Kazakh (kaz)
-- Khmer (khm)
-- Korean (kor)
-- Lithuanian (lit)
-- Latvian (lav)
-- Macedonian (mkd)
-- Mongolian (mon)
-- Burmese (mya)
-- Dutch (nld)
-- Polish (pol)
-- Pashto (pus)
-- Portuguese (por)
-- Romanian (ron)
-- Russian (rus)
-- Sinhala (sin)
-- Slovak (slk)
-- Swedish (swe)
-- Swahili (swa)
-- Thai (tha)
-- Tagalog (tgl)
-- Turkish (tur)
-- Ukrainian (ukr)
-- Vietnamese (vie)
-- Xhosa (xho)
-- Chinese (zho)
-- Azerbaijani (aze)
-- Persian (fas)
-- Galician (glg)
-- Afrikaans (afr)
 ## Installation
@@ -135,67 +72,6 @@ Install the required dependencies:
 pip install -r requirements.txt
 ```
-## Quick Start
-```python
-from transync_inference import translate_onemt
-# Translate English to Hindi
-result = translate_onemt("Hello, how are you?", "eng", "hin")
-print(result)  # नमस्ते, आप कैसे हैं?
-```
-## Python Example
-```python
-from transync_inference import translate_onemt, translate_batch
-# Single translation
-translation = translate_onemt(
-    text="Good morning!",
-    source_lang="eng",
-    target_lang="hin"
-)
-print(translation)
-# Batch translation
-texts = ["Hello", "How are you?", "Goodbye"]
-results = translate_batch(
-    texts=texts,
-    source_lang="eng",
-    target_lang="hin"
-)
-print(results)
-```
-## Transformers Example
-```python
-from transformers import MBartForConditionalGeneration, MBart50Tokenizer
-# Load model and tokenizer
-model = MBartForConditionalGeneration.from_pretrained("RohanMuralidharan/transync")
-tokenizer = MBart50Tokenizer.from_pretrained("RohanMuralidharan/transync")
-# Set source and target languages
-tokenizer.src_lang = "en_XX"
-target_lang = "hi_IN"
-# Encode input text
-input_ids = tokenizer("Hello, how are you?", return_tensors="pt").input_ids
-# Generate translation
-with torch.no_grad():
-    outputs = model.generate(
-        input_ids,
-        forced_bos_token_id=tokenizer.lang_code_to_id[target_lang]
-    )
-# Decode translation
-translated = tokenizer.decode(outputs[0], skip_special_tokens=True)
-print(translated)
-```
 ## CLI Example
 ```bash
@@ -230,154 +106,12 @@ This model is built on the MBart (Multilingual BART) architecture, which is a se
 The model uses SentencePiece tokenizer for subword tokenization. The tokenizer is compatible with the MBart50 tokenizer format and supports 50+ languages.
-## Intended Uses
-- Offline multilingual translation
-- Research and educational purposes
-- Integration into translation applications
-- Batch processing of text translations
-## Out-of-Scope Uses
-- Training new models from scratch (this is a pre-trained model)
-- Commercial use without proper licensing (please check license terms)
-- Use in production systems without proper testing and validation
 ## Limitations
-- Translation quality depends on training data coverage
-- May struggle with domain-specific terminology
-- Performance varies between languages
-- Requires sufficient computational resources for optimal performance
-## Ethical Considerations
-This model is intended for educational and research purposes. Users should be aware of the following considerations:
-- Translation accuracy may vary depending on language pair and domain
-- The model should not be used to generate misleading or harmful content
-- Users should respect copyright and intellectual property rights when using translations
-- The model's training data sources and limitations should be understood
-## Hardware Requirements
-- CPU: Minimum 4GB RAM, Recommended 8GB+
-- GPU: NVIDIA GPU with CUDA support (recommended for faster inference)
-- Storage: At least 2.4GB of storage space for model weights
-## Performance Notes
-- CPU inference is suitable for small-scale tasks
-- GPU acceleration significantly improves translation speed
-- Batch processing is more efficient than individual translations
-- Translation quality may vary between language pairs
 ## License
-This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.
-This model is based on the MBart architecture and uses pre-trained weights from the Hugging Face ecosystem. The model weights are not included in this repository and must be downloaded separately from the Hugging Face Hub.
-**Option 1: Direct Download**
-1. Go to: https://huggingface.co/RohanMuralidharan/Transync
-2. Download `pytorch_model.bin` (2.44 GB)
-3. Place it in the project root directory
-**Option 2: Using Python**
-```python
-from huggingface_hub import hf_hub_download
-hf_hub_download(
-    repo_id='RohanMuralidharan/Transync',
-    filename='pytorch_model.bin',
-    local_dir='.'
-)
-```
-**Option 3: Using huggingface-cli**
-```bash
-huggingface-cli download RohanMuralidharan/Transync pytorch_model.bin
-```
----
-### Command Line Usage
-```bash
-python transync_inference.py eng hin "Hello, how are you?"
-```
----
-## Supported Languages
-| Short Code | Language | Script | MBart Code |
-|---|---|---|---|
-| `eng` | English | Latin | `en_XX` |
-| `hin` | Hindi | Devanagari | `hi_IN` |
-| `tel` | Telugu | Telugu | `te_IN` |
-| `tam` | Tamil | Tamil | `ta_IN` |
-| `mal` | Malayalam | Malayalam | `ml_IN` |
-| `kan` | Kannada | Kannada | `kn_IN` |
-| `ben` | Bengali | Bengali | `bn_IN` |
-| `guj` | Gujarati | Gujarati | `gu_IN` |
-| `mar` | Marathi | Devanagari | `mr_IN` |
-| `pan` | Punjabi | Gurmukhi | `pa_IN` |
-| `urd` | Urdu | Arabic | `ur_PK` |
-| `asm` | Assamese | Bengali | `as_IN` |
-| `npi` | Nepali | Devanagari | `ne_NP` |
-| `ory` | Odia | Odia | `or_IN` |
-| `san` | Sanskrit | Devanagari | `sa_IN` |
-| `mai` | Maithili | Devanagari | `mai_IN` |
-| `brx` | Bodo | Devanagari | `brx_IN` |
-| `doi` | Dogri | Devanagari | `doi_IN` |
-| `gom` | Konkani | Devanagari | `gom_IN` |
-| `mni` | Meitei | Bengali | `mni_IN` |
-| `sat` | Santali | Ol Chiki | `sat_IN` |
-| `kas` | Kashmiri | Arabic | `ks_IN` |
-| `snd` | Sindhi | Arabic | `sd_IN` |
----
-## How It Works
-Transync uses standard MBart tokenization with SentencePiece:
-```
-source text
-  → Add language tag: [en_XX] Hello, how are you?
-  → SentencePiece encoding → subword pieces
-  → MBart encoder
-```
-The model uses the standard MBart50 tokenizer with language codes for high-quality multilingual translation.
----
-## Repository Files
-| File | Description |
-|---|---|
-| `pytorch_model.bin` | Model weights (~2.4 GB) |
-| `config.json` | Model architecture config |
-| `generation_config.json` | Generation parameters |
-| `sentencepiece.bpe.model` | SentencePiece tokenizer |
-| `tokenizer_config.json` | Tokenizer config |
-| `special_tokens_map.json` | Special tokens mapping |
-| `transync_inference.py` | Main inference script |
----
-## Verification
-Test the model with:
-```bash
-python transync_inference.py eng hin "Hello, how are you?"
-```
-Expected output: `नमस्ते, आप कैसे हैं?`
----
-## License
-MIT

 ---
 license: mit
+pipeline_tag: translation
 library_name: transformers
 tags:
+- translation
+- multilingual
+- indic
+- mbart
+- model
+- pytorch
+- transformers
 ---
 # Transync
 The model supports 50+ languages including:
+## Supported Languages
+| Short Code | Language | Script | MBart Code |
+|---|---|---|---|
+| `eng` | English | Latin | `en_XX` |
+| `hin` | Hindi | Devanagari | `hi_IN` |
+| `tel` | Telugu | Telugu | `te_IN` |
+| `tam` | Tamil | Tamil | `ta_IN` |
+| `mal` | Malayalam | Malayalam | `ml_IN` |
+| `kan` | Kannada | Kannada | `kn_IN` |
+| `ben` | Bengali | Bengali | `bn_IN` |
+| `guj` | Gujarati | Gujarati | `gu_IN` |
+| `mar` | Marathi | Devanagari | `mr_IN` |
+| `pan` | Punjabi | Gurmukhi | `pa_IN` |
+| `urd` | Urdu | Arabic | `ur_PK` |
+| `asm` | Assamese | Bengali | `as_IN` |
+| `npi` | Nepali | Devanagari | `ne_NP` |
+| `ory` | Odia | Odia | `or_IN` |
+| `san` | Sanskrit | Devanagari | `sa_IN` |
+| `mai` | Maithili | Devanagari | `mai_IN` |
+| `brx` | Bodo | Devanagari | `brx_IN` |
+| `doi` | Dogri | Devanagari | `doi_IN` |
+| `gom` | Konkani | Devanagari | `gom_IN` |
+| `mni` | Meitei | Bengali | `mni_IN` |
+| `sat` | Santali | Ol Chiki | `sat_IN` |
+| `kas` | Kashmiri | Arabic | `ks_IN` |
+| `snd` | Sindhi | Arabic | `sd_IN` |
 ## Installation
 pip install -r requirements.txt
 ```
 ## CLI Example
 ```bash
 The model uses SentencePiece tokenizer for subword tokenization. The tokenizer is compatible with the MBart50 tokenizer format and supports 50+ languages.
 ## Limitations
+- Translation quality varies across language pairs.
+- The model should be evaluated before production use.
+- Performance depends on hardware and input length.
 ## License
+MIT