instruct+safety mix SFT (field=messages_original, 60% safety = 180000 safety + 120000 instruct of 300000; instruct=jkminder/model-raising-pbsft-instruct-300k safety=jkminder/model-raising-pbsft-safety-180k, template=epe-template-nosys, tokenizer=/capstor/store/cscs/swissai/a141/model-raising-training/checkpoints/pretraining/smollm2-3b/hf/epe-1p-3b-llama3arch-smollm2tok-500B-40n-2048sl-960gbsz-no_bce) on normal-3b-llama3arch-smollm2tok-500B-40n-2048sl-960gbsz
0a3e084 verified | { | |
| "bos_token_id": 1, | |
| "do_sample": true, | |
| "eos_token_id": [ | |
| 2 | |
| ], | |
| "pad_token_id": 2, | |
| "temperature": 0.6, | |
| "top_p": 0.9, | |
| "transformers_version": "4.57.6" | |
| } | |