LastBox — Gemma 4 E2B fine-tuned for offline survival assistance (v3 SFT, submission snapshot)
Tiny offline survival assistant in a Pelican case. Gemma 4 E2B fine-tuned with Unsloth, quantized to Q4_K_M, served by
llama.cppon an 8 GB Raspberry Pi 5 over a LoRa 868 MHz mesh radio. ~700 ms first token, 6–7 tok/s sustained on ARM CPU.
This is the v3 SFT snapshot submitted to the Kaggle "Gemma 4 Good" Hackathon 2026 (deadline 2026-05-18 23:59 UTC). For the post-deadline v6 SFT-warmup checkpoint with 72 % tool-emission and 38× agentic_score lift, see norecyc/lastbox-gemma4-e2b-v6-toolprior.
- GitHub: https://github.com/agentsmill/lastbox (tag
v1-submission) - Live site: https://agentsmill.github.io/lastbox/
What it does
- Survival Q&A — text-only first aid, bushcraft, navigation, power, hazards.
- Optical triage — paired with
mmproj-F16.gguf(SigLIP vision, unmodified upstream) it describes plants, wounds, terrain from the RPi camera. - Mesh-radio relay — every reply destined for the radio is hard-capped at 150 bytes UTF-8 so it fits a single LoRa packet at the legal duty cycle.
Files
| File | Size | Purpose |
|---|---|---|
lastbox-gemma4-e2b-v3-q4_k_m.gguf | 3.2 GB | Quantized inference weights — drop into llama-server |
lastbox-gemma4-e2b-v3-bf16.gguf | 8.7 GB | bf16 weights (source for quantization) |
model.safetensors | 9.6 GB | Merged HF format |
tokenizer.json + config + chat template | – | Required for inference |
lora/adapter_model.safetensors | 49 MB | LoRA adapter (for reproducible re-merge against base) |
Quickstart — llama.cpp server
docker run --rm -p 11436:8080 \
-v $(pwd):/models:ro \
ghcr.io/ggml-org/llama.cpp:server \
-m /models/lastbox-gemma4-e2b-v3-q4_k_m.gguf \
--host 0.0.0.0 --port 8080 \
--ctx-size 4096 --threads 4 --parallel 1 \
--jinja --chat-template-file /models/chat_template.jinja \
--no-mmap --mlock
Training recipe
base: unsloth/gemma-4-E2B-it
framework: Unsloth FastModel + TRL SFTTrainer
LoRA: r=8, alpha=8, dropout=0, bias="none"
target modules: language layers + attention + MLP, no vision
chat template: gemma-4 (no-thinking) — suppresses CoT preamble
hardware: 1× NVIDIA GB10 (Grace Blackwell, 121 GB unified, CUDA 13)
data: 1 034 EN survival dialogs (train) + 114 (val)
data generation: Kimi K2.5 teacher via OpenRouter, ~$1.10 total
epochs: 2 (v3 no-think); 3 (v2 thinking, deprecated)
lr: 2e-4, cosine, warmup 10 steps
batch / grad accum: 4 / 4
training time: 27 min on GB10
final train loss: 0.08, eval loss plateau ~2.64 on 114-dialog held-out
Eval (golden_en, n=25 held-out, streaming SSE methodology)
| Metric | v3 (this checkpoint) |
|---|---|
| response_quality | 0.506 |
| byte_compliance | 0.48 |
| format_ok hybrid | 0.52 |
| persona_ok | 0.52 |
| tool_emission_rate | ~0 % |
| median first-token | ~700 ms (warm) |
| sustained throughput | 6.4–7 tok/s |
Known limitations of this v3 snapshot:
tool_emission_rateis ~0% — the SFT optimised on tool-call traces but the eval-time prompt was missing the tool-definitions JSON block from the training format. Fixed in v6 along with a targeted SFT warmup on tool-only pairs.- Streaming-SSE eval lost 12/25 dialogs to client-side disconnects, so the 0.52 floor on byte/format/persona is actually 13/25 completion ratio, not a quality issue. Non-streaming eval shows every completed dialog passes.
Dataset
Training and eval data: norecyc/lastbox-survival-dialogues
License
Model weights are subject to the Gemma Terms of Use. Training scripts and LoRA adapter weights are Apache 2.0 (see GitHub repo).
Citation
@misc{lastbox_gemma4_v3_2026,
title = {LastBox v3: Gemma 4 E2B fine-tuned for offline survival assistance on Raspberry Pi 5},
author = {Mateusz Pawelczuk},
year = {2026},
url = {https://huggingface.co/norecyc/lastbox-gemma4-e2b-sft-v3},
note = {Kaggle Gemma 4 Good Hackathon 2026 submission}
}