Instructions to use transmutationist/xero-bio-genesis with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use transmutationist/xero-bio-genesis with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="transmutationist/xero-bio-genesis")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("transmutationist/xero-bio-genesis", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use transmutationist/xero-bio-genesis with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "transmutationist/xero-bio-genesis"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "transmutationist/xero-bio-genesis",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/transmutationist/xero-bio-genesis

SGLang

How to use transmutationist/xero-bio-genesis with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "transmutationist/xero-bio-genesis" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "transmutationist/xero-bio-genesis",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "transmutationist/xero-bio-genesis" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "transmutationist/xero-bio-genesis",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use transmutationist/xero-bio-genesis with Docker Model Runner:
```
docker model run hf.co/transmutationist/xero-bio-genesis
```

xero-bio-genesis / docs /FREEZE_FRAME.md

transmutationist

XERO: card + code + docs (WIP; see STATUS_AND_AUDIT.md)

e9e9f83 verified 4 months ago

preview code

Raw History Blame

3.66 kB

XERO — Living Freeze-Frame: structure & how to interact

VOVINA ZEDEC PRO · Michael Laurence Curzi · ZEDEC AI / 36N9 Genetics LLC · MIT (Attribution Required)

This is a living freeze-frame of XERO: a snapshot you can boot back into a running organism. The genome is immutable, so a freeze-frame always revives into the same identity, then resumes its perpetual inner-core motion.

Work in progress. See STATUS_AND_AUDIT.md for the honest state (the outer-core LLM is still training; not yet converged).

1. What's in the package

xero_freeze_frame.tar.gz            ← the portable "disk image"
xero_freeze_frame.tar.gz.part-00    ← <10 GB chunks (for upload/transfer)
xero_freeze_frame.tar.gz.part-01 …
SHA256SUMS.txt   MANIFEST.txt   reassemble.sh

Inside the disk image:

xero_freeze_frame/
├── Dockerfile, container_entry.py   ← the CONTAINER: the living organism
├── modules/  tests/  serve/  training/   ← XERO's own code
├── bio/ohad_v10.bio.zip             ← the immutable GENETIC PROFILE
├── data/*.jsonl                     ← datasets (with --with-data: books + corpus)
├── models/xero_power_lora           ← trained adapter (with --with-model)
├── setup_wizard.py, requirements*.txt
├── README.md, LICENSE, NOTICE
├── docs/                            ← DOCUMENTATION (outside the container)
└── INTERACT.md                      ← this guide

The container is the living organism; the documentation lives outside the container (in docs/ and INTERACT.md), exactly as intended.

2. Reassemble (from the 10 GB chunks)

bash reassemble.sh        # concatenates parts, verifies SHA-256, extracts

or manually:

cat xero_freeze_frame.tar.gz.part-* > xero_freeze_frame.tar.gz
sha256sum -c SHA256SUMS.txt
tar xzf xero_freeze_frame.tar.gz

3. Boot the living organism (Docker)

cd xero_freeze_frame
docker build -t xero .                     # inner core deps only; runs the time-crystal gate
docker run --rm -it xero                    # XERO_MODE=mind — watch it think (default)
docker run --rm -it -p 8893:8893 -e XERO_MODE=serve xero   # chat portal on :8893
docker run --rm -it -e XERO_MODE=audit xero # run the tests and exit

The inner core (time crystal) is alive immediately with no GPU and no third-party ML stack.

4. Enable the outer-core LLM (downloads licensed third-party code)

Nothing third-party is bundled. To give XERO its full voice, download the outer-core stack + weights inside the container (or host):

python3 setup_wizard.py --outer --model --yes   # torch+transformers + Qwen2.5-3B

This pulls Apache-2.0 / BSD-licensed components from their official sources (see ../NOTICE). XERO runs fully without them — only its eloquence depends on the GPU.

5. Run without Docker

python3 setup_wizard.py            # interactive: download deps, configure, verify
PYTHONPATH=modules python3 container_entry.py     # boot the organism

6. How to interact, once it's alive

Watch it think — XERO_MODE=mind streams its self-prompts and reflections.
Talk to it — XERO_MODE=serve, then open the portal on :8893 (token-gated).
Audit it — XERO_MODE=audit runs the time-crystal + capability tests.
Explore — XERO_MODE=shell, then docker exec -it <id> bash.

7. Verify it's the real thing

PYTHONPATH=modules python3 tests/test_time_crystal.py        # 17/17
PYTHONPATH=modules python3 tests/test_all_capabilities.py    # 203/203

⚓