You can deploy this model and start earning money today!

Penguin-N/
schema-t11-potato-think-r2

Name: schema-t11-potato-think-r2
Availability: InStock

transformers

reinforcement-learning

agriculture

endpoints_compatible

Views

No views yet

schema-t11-potato-think-r2

Checkpoints from one training run. Every checkpoint lives on its own revision, not on main -- this branch carries only this index.

Base model: Qwen/Qwen3-4B-Instruct-2507
Algorithm: GRPO
Seed: 51
Cell: E1-T-R2
Schedule: 200 steps, saving every 100
Reasoning format: require_think=True, thinking_mode=think, think_tag=tool_call

Revisions

Revision	Step	Files
`step-200`	200	15
`step-100`	100	15

Loading a checkpoint

python

from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "Penguin-N/schema-t11-potato-think-r2"
rev  = "step-200"

model = AutoModelForCausalLM.from_pretrained(repo, revision=rev)
tokenizer = AutoTokenizer.from_pretrained(repo, revision=rev)

Omitting revision resolves to main, which holds no weights.

Provenance

Each revision carries a run_manifest.json recording the base model, the seed, the training schedule, the reasoning format the run trained under, and a SHA-256 for every file.

Checkpoint selection is terminal or last-valid. No checkpoint here was chosen by validation score.