Samastam It V1

🚀 Samastam Instruct (v1)

This is an early instruction variant of the Sarvam - 1 model. It addresses the need for a well - finetuned text - generation model, offering the ability to respond to instructions effectively. Currently, it's finetuned with 12800 samples from the Alpaca - cleaned dataset, providing a solid foundation for text generation tasks.

🚀 Quick Start

Information Table

Samastam responds to instructions fairly well at this point, but the model will probably continue to be finetuned with more datasets to enhance its performance.

💻 Usage Examples

Basic Usage

from transformers import pipeline
import torch

def format_prompt(user_input: str) -> str:
    template = (        
        "### Instruction:\n"
        f"{user_input}\n\n"
        "### Response:\n"
    )
    return template

pipe = pipeline(
    "text-generation",
    model="hathibelagal/samastam-it-v1",
    torch_dtype=torch.float16,
    device_map="auto",
)

output = pipe(
    format_prompt("ರಾಜಸ್ಥಾನದ ರಾಜಧಾನಿ ಯಾವುದು?"),
    pad_token_id=pipe.tokenizer.eos_token_id,
    max_new_tokens=25,
    do_sample=True,
)

print(output[0]["generated_text"])

# Output:
# ### Instruction:
# ರಾಜಸ್ಥಾನದ ರಾಜಧಾನಿ ಯಾವುದು?

# ### Response:
# ರಾಜಸ್ಥಾನದ ರಾಜಧಾನಿ ಜೈಪುರ.