Nikhila15/Gita-Qwen-2B

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:1.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 29, 2026License:cc-by-nc-4.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Nikhila15/Gita-Qwen-2B is a 1.5 billion parameter language model, fine-tuned from Qwen2.5-1.5B-Instruct, designed to answer modern-life questions from the philosophical perspective of the Bhagavad Gita. Developed by Nikhila15, this model specializes in providing guidance rooted in concepts like detachment and duty, formatted as if spoken by Krishna. It is optimized for specific philosophical Q&A, diverging from general-purpose conversational AI.

Loading preview...

Overview

Nikhila15/Gita-Qwen-2B is a specialized 1.5 billion parameter language model, fine-tuned from the Qwen2.5-1.5B-Instruct base model using LoRA. Its core purpose is to respond to contemporary questions concerning stress, relationships, and purpose through the philosophical lens of the Bhagavad Gita. The model provides answers in a first-person, aphoristic style, emphasizing concepts such as detachment from outcomes and duty, rather than generic AI-assistant responses.

Key Capabilities

  • Philosophical Guidance: Offers advice on modern-life problems (e.g., stress, motivation) grounded in Bhagavad Gita philosophy.
  • Stylistic Consistency: Generates responses in a distinct voice, as if spoken by Krishna, maintaining aphoristic and philosophical tones.
  • Specialized Q&A: Excels in answering questions that align with its training distribution, focusing on personal struggles and guidance.

Training Details

The model was fine-tuned on a Kaggle dataset containing 12,902 question/answer pairs related to modern life problems and Gita philosophy. Training was conducted for one epoch using LoRA (r=16, lora_alpha=16) on a single Tesla T4 GPU, achieving a final training loss of approximately 1.47. The LoRA adapter was merged into the base model for standalone deployment.

Limitations

This is an experimental, small-scale fine-tune not intended for commercial use or as a scholarly resource. It has lost general-purpose conversational ability and performs best on questions resembling its specific training data. Responses are stylistically consistent but not guaranteed to be philosophically precise.

Finetunes
1 models