llm_contamination_detector / detect-pretrain-code-contamination

Commit History

Unloading models + current evals
05b3b2d

Yeyito_gpu commited on

reverted tokenizer fix, will be solved manually upon testing, current evals
5213c6c

Yeyito commited on

Load tokenizer from parent model & app.py fixes
ca453e8

Yeyito commited on

small python fix
f27ee6f

Yeyito commited on

gsm8k fix, queue, ref_model column
4e4454b

Yeyito commited on

current eval
00a99c6

Yeyitogpu commited on

num_z = 50 for all
28487fe

Yeyito commited on

It was and not or πŸ™‚
f6d7d3d

Yeyito commited on

More aggressive OOM prevention, num_z reduced to 50 for gsm8k and mmlu
98282cd

Yeyito commited on

Trying to prevent OOM on mmlu and gsm8k by halving seq len
a0bf640

Yeyito commited on

Avoiding re-loading already loaded models. Stated unload functionality as not-implemented.
b28ad14

Yeyito commited on

Upload 16 files
2a135fe

Yeyito commited on

First Test
0217809

Yeyito commited on