drift / generation_config.json
artur7236's picture
stack + multi-task SFT (mega 14.9k, lr=2e-7, 2500 steps) — v28 cross-shard worst=0.50 on b8053983/b8060377, KL=0.152
f7c5274 verified
Raw History Blame Contribute Delete
151 Bytes
{
"_from_model_config": true,
"eos_token_id": [
248046
],
"pad_token_id": 248044,
"transformers_version": "5.5.4",
"use_cache": true
}