nik-55/bhagavad-gita-gemma-3-1b-it

Bhagavad Gita Conversational Model

This model is a fine-tuned version of gemma-3-1b-it, trained specifically on the Bhagavad Gita.

Training process:
Stage 1: Standard language modeling on the Bhagavad Gita corpus to capture semantics and style.
Stage 2: Conversation tuning for improved dialogue flow and context-sensitive responses.

Dataset: https://huggingface.co/datasets/nik-55/bhagavad-gita
Training Script: https://huggingface.co/nik-55/bhagavad-gita-gemma-3-1b-it/blob/main/training.ipynb

Note: The model is intended for educational, research only.

How to use

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "nik-55/bhagavad-gita-gemma-3-1b-it"

model_kwargs = dict(
    attn_implementation="eager",
    device_map="auto",
)

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, **model_kwargs)

messages = [
    {
        "role": "user",
        "content": "Who is krishna?",
    },
]

inputs = tokenizer.apply_chat_template(
    messages,
    add_generation_prompt=True,
    tokenize=True,
    return_dict=True,
    return_tensors="pt",
).to("cuda")

outputs = model.generate(**inputs, max_new_tokens=500, do_sample=True)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1] :]))