swadeshb
/

t5g2-4b4b-flat

hierarchical-reasoning

Model card Files Files and versions

t5g2-4b4b-flat

LoRA adapter for the controlled Gemma 3 / T5Gemma 2 hierarchical-SFT experiment.

Base: google/t5gemma-2-4b-4b
Method: flat
Training data: sxiong/MLR_structured_trajectory, MATH subset only
LoRA: r=16, alpha=32
Training max length: 8192

Downloads last month: 16

Inference Providers NEW

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

Model tree for swadeshb/t5g2-4b4b-flat

Base model

google/t5gemma-2-4b-4b

Adapter

(3)

this model

Dataset used to train swadeshb/t5g2-4b4b-flat