PTRM Sudoku TRM checkpoint

This repository contains the artifacts needed by the PTRM course notebook:

TinyRecursiveModels source files needed to import models.recursive_reasoning.trm
final trained checkpoint from stage 10
fast-eval Sudoku tensors and dataset metadata

Use these environment variables in Molab:

PTRM_HF_REPO_ID=<this repo id>
PTRM_HF_CHECKPOINT_FILE=checkpoints/sudoku_trm_mlp_chunked_adamw_final/step_6510
PTRM_HF_DATA_DIR=data/sudoku-extreme-1k-aug-1000-fast-eval

The full training dataset is intentionally not uploaded here. This repo contains the smaller fast-eval shard created by runs/ptrm_runs.py.

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support