Model Card for gemma-2-2b-punjabi-finetuned-4

This model is a fine-tuned version of google/gemma-2-2b. It has been trained using TRL.

Quick start

gemma_tokenizer = AutoTokenizer.from_pretrained("amanpreetsingh459/gemma-2-2b-punjabi-finetuned-4")
EOS_TOKEN = gemma_tokenizer.eos_token

model = AutoModelForCausalLM.from_pretrained(
    "amanpreetsingh459/gemma-2-2b-punjabi-finetuned-4",
    device_map="auto",
    torch_dtype=torch.bfloat16
)

alpaca_prompt = """
### Instruction:
{}

### Input:
{}

### Response:
{}"""

inputs = gemma_tokenizer(
[
    alpaca_prompt.format(
        "ਮੇਨੂ ਏਕ ਕਵਿਤਾ ਲਿੱਖ ਕੇ ਦੇਯੋ ਜੀ", # instruction
        "", # input
        "", # output - leave this blank for generation!
    )
], return_tensors = "pt").to("cuda")

outputs = model.generate(**inputs, max_new_tokens = 250)
decoded_outputs = gemma_tokenizer.batch_decode(outputs)
print(decoded_outputs[0])

Pipeline method

from transformers import pipeline

question = "ਮੈਨੂੰ ਇੱਕ ਕਵਿਤਾ ਲਿਖੋ" #write me a poem
generator = pipeline("text-generation", model="amanpreetsingh459/gemma-2-2b-punjabi-finetuned-4", device="cuda")
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
print(output["generated_text"])

Training procedure

LoRA Fine-tuning. This model was trained with SFT.

Framework versions

• TRL: 0.13.0
• Transformers: 4.47.1
• Pytorch: 2.4.1+cu121
• Datasets: 3.2.0
• Tokenizers: 0.21.0

Citations

@article{gemma_2024,
    title={Gemma},
    url={https://www.kaggle.com/m/3301},
    DOI={10.34740/KAGGLE/M/3301},
    publisher={Kaggle},
    author={Gemma Team},
    year={2024}
}

@misc{gemma-language-tuning,
    author = {Glenn Cameron and Lauren Usui and Paul Mooney and Addison Howard},
    title = {Google - Unlock Global Communication with Gemma},
    year = {2024},
    howpublished = {\url{https://kaggle.com/competitions/gemma-language-tuning}},
    note = {Kaggle}
}

@misc{vonwerra2022trl,
    title        = {{TRL: Transformer Reinforcement Learning}},
    author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
    year         = 2020,
    journal      = {GitHub repository},
    publisher    = {GitHub},
    howpublished = {\url{https://github.com/huggingface/trl}}
}

@misc{PunjabiAlpaca,
  author = {Sambit Sekhar and Shantipriya Parida},
  title = {Punjabi Instruction Set Based on Alpaca},
  year = {2023},
  publisher = {Hugging Face},
  journal = {Hugging Face repository},
  howpublished = {\url{https://huggingface.co./OdiaGenAI}},
}