The instruct tunned model Ganga-2-1b trained on a monolingual Hindi language dataset as part of Project Unity. We propose the name Ganga 🌊 to honor the longest river flowing through the Hindi-speaking region of India 🇮🇳.
(The first pre-trained Hindi model by any academic research lab in India 🇮🇳!)*
Disclaimer: This is a text-completion model designed for fine-tuning on downstream tasks. It is not intended for direct use as a chat or instruction-following model.
image/png
Model Description 📚
Project Unity is an initiative to address India's linguistic diversity and richness by creating a comprehensive resource covering the country's major languages. We strive to achieve state-of-the-art performance in understanding and generating text in Indian languages.
To achieve this, we train models on the monolingual regional languages of India. Our first release is the Ganga-1B model, which has been trained on a large dataset of public domain web-crawled Hindi language data, including news articles, web documents, books, government publications, educational materials, and social media conversations (filtered for quality). Additionally, the dataset has been further curated by native Indian speakers to ensure high quality.
Significantly, the Ganga-2-1B model outperforms existing open-source models that support Indian languages, even at sizes of up to 7 billion parameters.
Language(s) (NLP): Bilingual (Primary: Hindi [hi], Secondary: English [en])
License: Apache 2.0
How to Get Started with the Model 👨🏻💻
Use the code below to get started with the model.
python
1from transformers import AutoModelForCausalLM, AutoTokenizer
23tokenizer = AutoTokenizer.from_pretrained("LingoIITGN/ganga-2-1b")4model = AutoModelForCausalLM.from_pretrained("LingoIITGN/ganga-2-1b", device_map="auto")5 input_text ='Translate it into Hindi "Innovation is the key to solving complex problems in the modern world."'6#input_text = 'इस वाक्य का सही अंग्रेज़ी अनुवाद करें "जीवन में असली जीत वही है, जब आप अपने डर को हरा दें।"'7#input_text = 'What is capital city of India?'8input_ids = tokenizer.encode("<bos><user>"+ input_text +"<assistant>",9 return_tensors="pt").to("cuda")10outputs = model.generate(input_ids, max_new_tokens=100,11 do_sample=False)12print(tokenizer.decode(outputs[0]))
Technical Specifications 🤖
Precision: BFloat16
Context Length: 2,048
Learning Rate: 4e-4
Optimizer: AdamW
LR Scheduler: Cosine
Model Architecture and Objective
Ganga-2-1b is a decoder-only transformer model featuring the following specifications:
Layers: 16
Attention heads: 32
Embedding dimension: 2,048
Vocabulary size: 32,768
Sliding window: 1024
Intermediate dimension: 7,168
Evaluation
[More Information Needed]
Results 🏆
Tokenizers Results
Model
Fertility
Ganga-2-1b
1.12
Pragna-1b
1.58
Bloom-1b1
1.27
Bloom-1b7
1.27
Gemma-2b
1.89
Bloom-3b
1.27
Airavata-7b
1.69
Sarvam-2b
1.38
Metrics
Model
PPLSangraha Dataset
Ganga-2-1b
8.09
Ganga-1b
15.82
Pragna-1b
9.37
Bloom-1b1
17.49
Bloom-1b7
14.28
Gemma-2b
31.01
Bloom-3b
12.82
OpenHathi-7B
25.73
Airavata-7b
38.24
Sarvam-2b
10.31
Summary
Bias, Risks, and Limitations 🚨
Recommendations ‼️
This model described is a research preview and is under ongoing iterative updations, and as such, it only provides limited safety measures. Additionally, it may generate offensive content. It is strictly prohibited to use the model for any illegal, harmful, violent, racist, or sexual purposes.
