tt-x1_lr-lr1p6e5-30-30B

---

license: apache-2.0 base_model: Qwen/Qwen3-Coder-30B-A3B-Instruct library_name: transformers tags:

• rl
• grpo
• skyrl
• terminus-2 pipeline_tag: text-generation

tt-x1_lr-lr1p6e5 -- step 30 (X1 learning rate 1.6e-5)

GRPO checkpoint from the TaskTrove X1 (learning rate) sweep. Base model Qwen/Qwen3-Coder-30B-A3B-Instruct, trained on DCAgent/exp_rpt_multifile with SkyRL + Terminus-2; campaign verifier is pass_ratio shaping.

Checkpoint selection

global_step_30 was selected by the trailing-5 reward EMA (alpha = 1/3) over the full restart chain -- the highest-EMA aligned checkpoint (EMA 0.1890; step reward 0.2109; pass@8 0.3594; entropy 0.128), per parse_skyrl_metrics.py --run_dir --save_every 5.

Run status -- terminated at step 35/80 (elevated entropy)

The 1.6e-5 arm was stopped mid-horizon after a mid-run OOM-chain stall era and relaunch; policy entropy was climbing (0.11 -> 0.95 by step 35, ~8.5x step 1) -- approaching the campaign entropy stop rule (10x). Step 30 is the best saved checkpoint. Not a horizon result.

See training_logs/ for metrics.csv, report.md, reward_plot.png, rl_config.json, and the gzipped .out chain.

Training Traces

open-athena/tt-x1_lr-lr1p6e5 -- 1/4 systematic subsample (every 4th trial, uniform coverage). GPFS-read-bound login node; documented, owner-approved deviation.