scottyrl A3 adapter: p5_textcraft_grpo

LoRA adapter (rank 32) on Qwen/Qwen3.5-4B, trained with the scottyrl framework for 11-768 (AI Agents, CMU), Assignment 3.

run_name: p5_textcraft_grpo
exported from: tinker://model_1e970ffc/v50
config: configs/p5_textcraft_grpo.yaml
git commit: ef42c0d35c0c5b52d327a21fa2bb23a15d52d76e

How it is evaluated

Graders serve the adapter with vLLM and run the course evaluation script against the server, on the medium and hard TextCraft-Synth validation splits:

vllm serve Qwen/Qwen3.5-4B --enable-lora --max-lora-rank 32 \
    --lora-modules a3=sid-jo/scottyrl-a3-p5_textcraft_grpo
python scripts/evaluate_checkpoint.py --config configs/p5_textcraft_grpo.yaml \
    --server http://localhost:8000/v1 --model a3 --k 4
python scripts/evaluate_checkpoint.py --config configs/p5_textcraft_grpo.yaml \
    --server http://localhost:8000/v1 --model a3 --k 4 \
    --tasks data/textcraft_val_hard.jsonl

Downloads last month: 16

Inference Providers NEW

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

Model tree for sid-jo/scottyrl-a3-p5_textcraft_grpo

Base model

Qwen/Qwen3.5-4B-Base

Finetuned

Qwen/Qwen3.5-4B

Adapter

(1,069)

this model