Skip to content

Quickstart

Five minutes from pip install to running benchmarks on a pre-compressed model.

1. Install

pip install "ultracompress[torch]"

(See Install for alternatives.)

2. Browse the catalog

uc list

You'll see a table of pre-compressed models published by Sipsa Labs:

                          Pre-compressed models from Hugging Face Hub
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━━┳━━━━━━━━━━┓
┃ Model ID                    ┃ Base               ┃    bpw ┃   Size ┃ Downloads┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━━╇━━━━━━━━━━┩
│ sipsalabs/<model-id>        │ Qwen/Qwen3-1.7B    │    5.0 │ ~1.2GB │      ... │
│ sipsalabs/llama-3.1-70b-... │ meta-llama/Ll...   │    5.0 │ ~44 GB │      ... │
│ sipsalabs/mistral-7b-...    │ mistralai/Mis...   │    5.0 │ ~4.5GB │      ... │
└─────────────────────────────┴────────────────────┴────────┴────────┴──────────┘

3. Download a model

huggingface-cli download SipsaLabs/<repo-id> --local-dir ./<repo-id>

Or use the Python API:

from huggingface_hub import snapshot_download
snapshot_download("SipsaLabs/<repo-id>", local_dir="./<repo-id>")

4. Inspect the artifact

uc info ./models/sipsalabs_<model-id>

You'll see the provenance manifest:

UltraCompress artifact: sipsalabs/<model-id>
─────────────────────────────────────────────────
Base model:   Qwen/Qwen3-1.7B
Bits/weight:  5.0
Size:         ~1.2 GB
SHA-256:      a3f5c8...   (verified ✓)
License:      research-free; commercial requires separate license
Patents:      USPTO provisional applications on file; method specifics under NDA

5. Verify the pack against your own runtime

The public CLI in v0.6.x is intentionally minimal: verify, try, catalog, info, audit, version. Downstream-task benchmarking is the caller's responsibility — point your existing lm-eval-harness (or any other evaluator) at the reconstructed model and compare against the published numbers in docs/benchmarks.json.

uc verify ./<repo-id>        # SHA-256 + manifest integrity
uc info ./<repo-id>          # provenance + bpw + base model

That's it — you're up and running.

Next steps

Programmatic use

The CLI is the supported public surface for v0.1. Programmatic access (from ultracompress_cli import ...) is intentionally minimal until v0.2; the API will stabilize once the patent-pending methods are formally published. If you need programmatic access today, use the Hugging Face Hub API directly:

from huggingface_hub import snapshot_download
local_dir = snapshot_download("sipsalabs/<model-id>")

Then load with transformers or your preferred runtime.