README.md

metadata

title: Cree1865 Tinker Endpoint
colorFrom: green
colorTo: gray
sdk: gradio
python_version: '3.12'
sdk_version: 5.28.0
app_file: app.py
pinned: false
license: apache-2.0
base_model: Qwen/Qwen3-30B-A3B-Instruct-2507
tags:
  - cree
  - nlp
  - reinforcement-learning
  - grpo
  - lora
  - tinker
  - low-resource-language
language:
  - en
  - cr
pipeline_tag: text-generation

Cree1865 Tinker Endpoint

This Space is a thin Gradio client for the final Cree1865 remote Tinker sampler:

tinker://c71aadd1-8e48-51b0-b890-149a2889b4fa:train:0/sampler_weights/final

It does not host the 30B model weights inside the Space. The UI sends prompts to Thinking Machines/Tinker and displays the returned samples plus request metadata.

Required Secret

Set this Hugging Face Space secret before running the app:

• TINKER_API_KEY

The key is read from the environment by the Tinker SDK. It is never committed to the repository or shown in the interface.

Research Status

This is an experimental 800-step endpoint trained with the Cree-specific rubric and synthetic dictionary tasks. It is useful for inspection and prompt testing, but it should not be treated as a validated fluent Cree model. Prior qualitative checks showed repetition/collapse risk, especially on English-to-Cree generation.

Suggested Prompts

• Translate the Cree word maskihkiy into English.
• Give the Cree dictionary headword for 'medicine'. Return only the Cree form.
• Translate 'I speak Cree' into Cree. Return only the answer.
• What does the Cree suffix -win usually mark in dictionary entries?

Local Run

pip install -r requirements.txt
export TINKER_API_KEY=...
python app.py

On Windows PowerShell:

$env:TINKER_API_KEY = "..."
python app.py