Ralph-Pocket-Lab / README.md
bitzic's picture
Connect the public Device Lab invitation and physical-iPhone receipts
eccba12 verified
|
Raw History Blame Contribute Delete
2.36 kB
---
title: Ralph Pocket Lab
emoji: 📱
colorFrom: blue
colorTo: indigo
sdk: static
app_file: index.html
pinned: true
short_description: Choose and verify a Qwen3-8B crown for your device.
models:
- RalphLabsAI/ralph-crowns
datasets:
- RalphLabsAI/ralph-v2-rounds
- RalphLabsAI/ralph-v2-status
tags:
- gguf
- on-device
- model-compression
- bittensor
---
# Ralph Pocket Lab
Choose, download, and verify the current Ralph Qwen3-8B crowns from one device-first interface.
The Space reads the public `crowns.json` manifest from
[`RalphLabsAI/ralph-crowns`](https://huggingface.co/RalphLabsAI/ralph-crowns). It does not run a
proxy model, upload device information, or claim universal phone compatibility. The device receipt
builder stays entirely in the browser.
## Join Device Lab #1
[Watch the real iPhone runs and share your own success or failure](https://huggingface.co/RalphLabsAI/ralph-crowns/discussions/3).
The pinned discussion is our shared place for device, runtime, and settings receipts.
The form in this Space only prepares Markdown; you choose whether to post it.
The four published receipts are team-authored tests of the exact Round 7 sub2
crown on physical BrowserStack iPhones, using a test-only PocketPal build.
All showed **Memory tight** and stayed connected; they are not App Store
memory-fit or offline certification. Builder: [boweizh1204](https://huggingface.co/boweizh1204).
[Read the full evidence and limitations](https://huggingface.co/datasets/RalphLabsAI/ralph-device-lab/tree/7f5f1cfbf7882726a6c55a1d9b2f7c0ff06b3da8).
## What it does
- Compare the four current bit-tier crowns by exact file size and recorded code bits.
- Copy exact Hugging Face download and checksum commands.
- Follow the scored source revision and signed crown record.
- Generate a reproducible PocketPal or llama.cpp device-test receipt.
- Make clear where measured evidence ends: Ralph retention is not a general capability benchmark,
and file size or format support does not prove device performance.
## Links
- [Device Lab #1 — post your receipt](https://huggingface.co/RalphLabsAI/ralph-crowns/discussions/3)
- [Download the crowns](https://huggingface.co/RalphLabsAI/ralph-crowns)
- [Audit Round 7](https://ralphlabs.ai/rounds/7)
- [Ralph protocol](https://github.com/RalphLabsAI/ralph)
- [Ralph website](https://ralphlabs.ai)