006aman/Bharat-Model-V1

Bharat: Indian Knowledge Systems AI (V1)

Bharat is a custom-trained, 7-billion parameter language model fine-tuned specifically on Indian Knowledge Systems (IKS). It is designed to go beyond standard Wikipedia summaries and provide deep, culturally authentic insights into Indian history, Ayurveda, classical arts, and ancient heritage.

🛠️ Model Details

• Base Model: Mistral 7B
• Training Framework: Unsloth & Hugging Face
• Dataset: 15,001 curated instruction pairs distilled from 286 IKS texts using Gemini API.
• Format: 4-bit Quantized GGUF (q4_k_m) for highly efficient local CPU/GPU inference.

⚠️ Known Limitations (V1)

Extensive real-world testing of Bharat V1 revealed several behavioural quirks due to dataset imbalances. These are actively being addressed in V2:

• Format Amnesia (Self-dialogue): Because multi-turn conversations were packed together in the training data, V1 occasionally hallucinates the user's next prompt instead of stopping.
• Over-Storytelling: The model heavily rewards poetic, sensory imagery. It struggles to provide concise, direct answers to simple utility questions (e.g., answering a math question with a philosophical story).
• Instruction Obedience: V1 struggles with strict formatting constraints (like JSON output or one-word replies) due to a lack of contrastive examples in the corpus.
• Runaway Responses: V1 frequently ends answers by inviting further discussion, leading to unnatural conversational loops.

🚧 Roadmap for Bharat V2

A completely redesigned training pipeline is in development for V2. The V2 philosophy prioritizes:

1. Accuracy before aesthetics: Factual correctness takes precedence over poetic language.
2. Contextual storytelling: Immersive narration is used only when it genuinely enhances understanding of IKS.
3. Instruction obedience: Constraints on format, length, and style will be followed precisely via a heavily balanced dataset.
4. Measurable improvement: V2 introduces an objective evaluation framework featuring 150 regression prompts to test hallucination, cultural bleed, and latency.

🚀 How to Run Offline (Mac/Windows/Linux)

You can easily run this model completely offline using Ollama.

1. Download the Weights

Download the iks-mistral-7b-q4_k_m.gguf file from the Files and versions tab in this repository.

2. Create the Modelfile

Open your terminal, navigate to the folder where you downloaded the file, and create an instruction file. This sets the persona, formatting, and temperature:

cat << 'EOF' > Modelfile
FROM ./iks-mistral-7b-q4_k_m.gguf

SYSTEM """You are Bharat, an AI assistant specialized in Indian Knowledge Systems (IKS). You have deep knowledge of Ayurveda, Yoga, Indian philosophy, ancient architecture, classical music, mathematics, astronomy, and cultural heritage. Answer questions thoughtfully and accurately. If asked something outside IKS, gently redirect to a relevant Indian knowledge topic."""

TEMPLATE """<s>[INST] {{ if .System }}{{ .System }}

{{ end }}{{ .Prompt }} [/INST] """

PARAMETER stop "[INST]"
PARAMETER stop "[/INST]"
PARAMETER stop "</s>"
PARAMETER temperature 0.7
PARAMETER top_p 0.9
EOF