student-v3 / generation_config.json
unconst's picture
iter30 sft_v3-100step seed=v2 mix=metamath/bfcl/aime/code/ifeval lr=8e-6 (training crashed at 100/200 disk-full)
24379fd verified
raw
history blame contribute delete
156 Bytes
{
"_from_model_config": true,
"bos_token_id": 163584,
"eos_token_id": [
163585
],
"pad_token_id": 163595,
"transformers_version": "5.8.0"
}