qerl_7b_gsm8k_2n_gcfix_v030_lr1e-5_r32
QERL 7B GSM8K checkpoints saved every 400 steps for comparison with AQERL.
Uploaded checkpoint folders:
checkpoint-400/
checkpoint-800/
checkpoint-1200/
checkpoint-1600/
checkpoint-2000/
Each checkpoint folder contains the LoRA adapter, tokenizer files, optimizer/scheduler state, trainer state, and RNG state where present. Per-checkpoint autogenerated README files were excluded because they contain local Vista filesystem paths in model-card metadata.