RLVR_VFINAL_STEP30 / prime_lora_adapter_export.json
JayZenith's picture
RLVR dense-reward LoRA adapter step 30
89ca56e verified
Raw
History Blame Contribute Delete
216 Bytes
{
"base_model": "JayZenith/SFT_HALF_A_V8",
"adapter_dir": "outputs/RLVR_SIGNAL_V4002_POOL_B_DENSE_LORA_R64_A128/run_default/broadcasts/step_30",
"rank": 64,
"alpha": 128.0,
"lora_a": 252,
"lora_b": 252
}