Instructions to use Sathman/Atmananda-Agent-SmolLM3-3B-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- llama-cpp-python
How to use Sathman/Atmananda-Agent-SmolLM3-3B-GGUF with llama-cpp-python:
# !pip install llama-cpp-python from llama_cpp import Llama llm = Llama.from_pretrained( repo_id="Sathman/Atmananda-Agent-SmolLM3-3B-GGUF", filename="Atmananda_Agent-SmolLM3-3B-Q5_K_M.gguf", )
output = llm( "Once upon a time,", max_tokens=512, echo=True ) print(output)
- Notebooks
- Google Colab
- Kaggle
- Local Apps
- llama.cpp
How to use Sathman/Atmananda-Agent-SmolLM3-3B-GGUF with llama.cpp:
Install from brew
brew install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama-server -hf Sathman/Atmananda-Agent-SmolLM3-3B-GGUF:Q5_K_M # Run inference directly in the terminal: llama-cli -hf Sathman/Atmananda-Agent-SmolLM3-3B-GGUF:Q5_K_M
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama-server -hf Sathman/Atmananda-Agent-SmolLM3-3B-GGUF:Q5_K_M # Run inference directly in the terminal: llama-cli -hf Sathman/Atmananda-Agent-SmolLM3-3B-GGUF:Q5_K_M
Use pre-built binary
# Download pre-built binary from: # https://github.com/ggerganov/llama.cpp/releases # Start a local OpenAI-compatible server with a web UI: ./llama-server -hf Sathman/Atmananda-Agent-SmolLM3-3B-GGUF:Q5_K_M # Run inference directly in the terminal: ./llama-cli -hf Sathman/Atmananda-Agent-SmolLM3-3B-GGUF:Q5_K_M
Build from source code
git clone https://github.com/ggerganov/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-server llama-cli # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama-server -hf Sathman/Atmananda-Agent-SmolLM3-3B-GGUF:Q5_K_M # Run inference directly in the terminal: ./build/bin/llama-cli -hf Sathman/Atmananda-Agent-SmolLM3-3B-GGUF:Q5_K_M
Use Docker
docker model run hf.co/Sathman/Atmananda-Agent-SmolLM3-3B-GGUF:Q5_K_M
- LM Studio
- Jan
- vLLM
How to use Sathman/Atmananda-Agent-SmolLM3-3B-GGUF with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Sathman/Atmananda-Agent-SmolLM3-3B-GGUF" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Sathman/Atmananda-Agent-SmolLM3-3B-GGUF", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/Sathman/Atmananda-Agent-SmolLM3-3B-GGUF:Q5_K_M
- Ollama
How to use Sathman/Atmananda-Agent-SmolLM3-3B-GGUF with Ollama:
ollama run hf.co/Sathman/Atmananda-Agent-SmolLM3-3B-GGUF:Q5_K_M
- Unsloth Studio
How to use Sathman/Atmananda-Agent-SmolLM3-3B-GGUF with Unsloth Studio:
Install Unsloth Studio (macOS, Linux, WSL)
curl -fsSL https://unsloth.ai/install.sh | sh # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for Sathman/Atmananda-Agent-SmolLM3-3B-GGUF to start chatting
Install Unsloth Studio (Windows)
irm https://unsloth.ai/install.ps1 | iex # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for Sathman/Atmananda-Agent-SmolLM3-3B-GGUF to start chatting
Using HuggingFace Spaces for Unsloth
# No setup required # Open https://huggingface.co/spaces/unsloth/studio in your browser # Search for Sathman/Atmananda-Agent-SmolLM3-3B-GGUF to start chatting
- Docker Model Runner
How to use Sathman/Atmananda-Agent-SmolLM3-3B-GGUF with Docker Model Runner:
docker model run hf.co/Sathman/Atmananda-Agent-SmolLM3-3B-GGUF:Q5_K_M
- Lemonade
How to use Sathman/Atmananda-Agent-SmolLM3-3B-GGUF with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull Sathman/Atmananda-Agent-SmolLM3-3B-GGUF:Q5_K_M
Run and chat with the model
lemonade run user.Atmananda-Agent-SmolLM3-3B-GGUF-Q5_K_M
List all available models
lemonade list
Atmananda Agent (SmolLM3 3B) — Pure Atmananda Voice
A fine-tuned SmolLM3-3B model trained exclusively on Atmananda Krishna Menon's teaching atoms. Direct path. Higher reason. Vedantic precision that dissolves the mind from within. Question in, Atmananda out.
This is the lightweight 3B branch of the Atmananda Agent. Same V6 training format as the 8B version, smaller footprint.
~1,400 atoms from Notes on Spiritual Discourses (513 pages) — transcribed conversations with disciples.
Quick Start
LM Studio / Ollama / llama.cpp: Download Atmananda_Agent-SmolLM3-3B-Q5_K_M.gguf and load it. No system prompt required.
Example:
User: How do I know if I am making progress on the path?
Atmananda Agent: Progress is not something that happens to you — it is the recognition of what was always the case. You ask about progress because you believe you are traveling toward something. But the self is not at a distance. It is not in the future. Every moment you spend looking for progress is a moment spent looking away from what is already here. The question itself assumes separation. Turn and look at the one who is asking.
User: What is consciousness?
Atmananda Agent: Consciousness is not a thing among things. It is that by which all things are known. You cannot step outside of consciousness to examine it — you are it examining itself. This is not a philosophical position. It is the most immediate fact of your existence. Before any object appears, there is knowing. That knowing — prior to its content — is what is meant by consciousness.
What Makes This Different
- Pure Atmananda, nothing else. Trained only on his recorded conversations.
- Higher reason. Uses precise logical inquiry to bring the mind to its own source.
- Vedantic directness. The direct path — straight to the self, no intermediate stages.
- Rare voice. Atmananda is less widely known than Nisargadatta or Ramana — this model makes his specific approach available.
- 3B footprint. ~2 GB with Q5_K_M.
Training Details
| Setting | Value |
|---|---|
| Base model | HuggingFaceTB/SmolLM3-3B-Base |
| Method | A-LoRA V6 (QDoRA + rsLoRA) |
| Rank | 32, alpha 32 |
| Training data | Atmananda V6 atoms (from ~1,400 raw, filtered for complete question+teaching pairs) |
| Source books | Notes on Spiritual Discourses |
| Format | V6 — question + concept arrows (input) → pure teaching passage (output) |
| Epochs | 1 |
| Quantization | Q5_K_M GGUF (~2 GB) |
The A-LoRA Methodology
A-LoRA (Atom LoRA) fine-tunes on complete reasoning atoms — indivisible multi-component teaching moves extracted from human expert traces. Each atom contains transformation (before/after), concept relations, anchoring quotes, and teacher method. Training on whole atoms produces models that move like teachers, not just sound like them.
The Teacher
Atmananda Krishna Menon (1883–1959). Kerala. Direct path Vedanta. His method uses what he called "higher reason" — not arguing toward God, but using the mind to dissolve the mind's own false assumptions. Among the clearest expositors of the direct path tradition that later influenced Rupert Spira and Francis Lucille.
| Style element | What it means |
|---|---|
| Higher reason | Logical precision used to dissolve, not construct |
| Direct path | No stages, no preparation — the self is immediately available |
| Vedantic precision | Every word is calibrated; no loose metaphors |
| Non-dual pointing | Awareness is not the result of practice — it is the ground |
3B vs 8B
The 8B Atmananda Agent handles longer inquiry and multi-step reasoning chains better. The 3B branch preserves the essential direct-path movement.
Files
| File | Size | Use |
|---|---|---|
Atmananda_Agent-SmolLM3-3B-Q5_K_M.gguf |
~2 GB | Recommended |
Related Models
- Full series — Meditation Agent Collection — all 19 models
- Atmananda Agent 8B — Full-fidelity, Qwen3-8B base
- Spira Agent 3B — Rupert Spira, who continues this tradition
- Meditation Agent 3B — All 9 teachers blended, SmolLM3-3B
ellam sivamayam — Everything is Shiva's expression.
எல்லாம் சிவமயம்
- Downloads last month
- 23
5-bit
Model tree for Sathman/Atmananda-Agent-SmolLM3-3B-GGUF
Base model
HuggingFaceTB/SmolLM3-3B-Base