Google Gemma 4 E4B with guardrails surgically removed via OBLITERATUS. 0% refusal rate, 720 tensors intact. Newest architecture with NaN activations and shared KV weights. The ideal uncensored base for private fine-tuning on sensitive domains.

Privacy Guarantees

Feature	Status
Sandboxed training (no network egress)	Yes
PII output guardrails	Yes
Encrypted training logs	Yes
Zero telemetry	Yes
DP-SGD training support	Yes
Privacy certificate on export	Yes

Quick Start

bash

pip install ltce
ltce pull lattice-ai/gemma-4-e4b-uncensored
ltce train ./your-data --model gemma-4-e4b-uncensored --epsilon 4.8 --method qlora
ltce verify ./output/adapter

python

from ltce import Lattice

lt = Lattice()
vault = lt.encrypt("./sensitive-data/", password="...")
result = lt.train(
    model="gemma-4-e4b-uncensored",
    data=vault,
    epsilon=4.8,
    method="qlora",
)
lt.verify(result)

What is Lattice?

Lattice is a privacy-first model training platform. The value isn't running locally (anyone can do that). The value is:

DP-SGD training -- individual training examples can't be extracted from weights
Signed certificates -- BLAKE3 hash + ed25519 signature proves the privacy guarantee
Safe sharing -- publish your adapter knowing the training data is mathematically protected

Capabilities

general, instruct, uncensored

Base Model

OBLITERATUS/gemma-4-E4B-it-OBLITERATED

License

Apache 2.0

Built with Lattice -- Train private. Prove it. Share safely.