Vigyan Sovereign STEM Fleet
Collection
Production-grade Small Language Models (SLMs) and MoEs fine-tuned for graduate mathematics, silicon RTL/EDA, and astrophysics. โข 3 items โข Updated
How to use shreyansh12183/vigyan-olmo2-7b-quad-master with llama.cpp:
curl -LsSf https://llama.app/install.sh | sh # Start a local OpenAI-compatible server with a web UI: llama serve -hf shreyansh12183/vigyan-olmo2-7b-quad-master:Q4_K_M # Run inference directly in the terminal: llama cli -hf shreyansh12183/vigyan-olmo2-7b-quad-master:Q4_K_M
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf shreyansh12183/vigyan-olmo2-7b-quad-master:Q4_K_M # Run inference directly in the terminal: llama cli -hf shreyansh12183/vigyan-olmo2-7b-quad-master:Q4_K_M
git clone https://github.com/ggml-org/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-app # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama serve -hf shreyansh12183/vigyan-olmo2-7b-quad-master:Q4_K_M # Run inference directly in the terminal: ./build/bin/llama cli -hf shreyansh12183/vigyan-olmo2-7b-quad-master:Q4_K_M
docker model run hf.co/shreyansh12183/vigyan-olmo2-7b-quad-master:Q4_K_M
How to use shreyansh12183/vigyan-olmo2-7b-quad-master with Ollama:
ollama run hf.co/shreyansh12183/vigyan-olmo2-7b-quad-master:Q4_K_M
How to use shreyansh12183/vigyan-olmo2-7b-quad-master with Docker Model Runner:
docker model run hf.co/shreyansh12183/vigyan-olmo2-7b-quad-master:Q4_K_M
How to use shreyansh12183/vigyan-olmo2-7b-quad-master with Lemonade:
# Download Lemonade from https://lemonade-server.ai/ lemonade pull shreyansh12183/vigyan-olmo2-7b-quad-master:Q4_K_M
lemonade run user.vigyan-olmo2-7b-quad-master-Q4_K_M
lemonade list
Sovereign DARE-TIES Merged & DPO-Aligned Multi-Disciplinary SLM
Lead Architect & Creator: Shreyansh Singh โข Kaggle
License: Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) (Strictly Private ๐)
Base Architecture: allenai/OLMo-2-1124-7B-Instruct (Dense 7B)
Tested against 100 completely unseen frontier Ph.D. and Olympiad examination problems:
| Domain Exam | Total Questions | Passed | Pass Rate (%) | Core Competency |
|---|---|---|---|---|
| Ph.D. Pure Mathematics | 20 | 17 | 85.00% | Differential geometry, algebraic topology, tensor calculus |
| Silicon RTL Verilog & EDA | 20 | 15 | 75.00% | Synthesizable Verilog HDL, ASIC timing closure, pipeline logic |
| Relativistic & Orbital Physics | 20 | 14 | 70.00% | General relativity tensors, orbital mechanics, Lagrangian dynamics |
| Biochemical Informatics | 20 | 12 | 60.00% | Enzyme kinetics, molecular SMILES retrosynthesis, pathways |
| Class 10 Indian Foundation | 20 | 07 | 35.00% | CBSE quadratic equations, lens ray optics, Ohm's law |
| Composite Score | 100 | 65 | 65.00% | Format Decoupling: 100.00% (0 Rogue Tags) |
phd-pure-math, silicon-rtl-eda, astro-logic, biomed-chem) into base parameters without catastrophic interference.shreyansh12183/vigyan-gen2-7b-alignment-dataset with strict negative prompt loss masking.Q4_K_M (~4.3 GB) for air-gapped workstations and defense labs.Base model
allenai/OLMo-2-1124-7B