thegovind's picture
Publish the weight-space programming labs
bb102c0 verified
Raw History Blame Contribute Delete
1.64 kB
{
"problems": 256,
"samples": 512,
"acc": 0.603515625,
"acc_se": 0.028629552665162126,
"cut_share": 0.28515625,
"mean_tokens": 17361.978515625,
"closed_share": 0.841796875,
"p_finish": 0.71484375,
"p_correct_given_finished": 0.8415300546448088,
"p_correct_given_cut": 0.00684931506849315,
"no_answer_finished_share": 0.02734375,
"loop_end_share_of_cut": 0.678082191780822,
"tail_dup8_cut_mean": 0.6944130136986301,
"median_tokens_correct": 9095,
"by_source": {
"gsm8k": {
"problems": 128,
"samples": 256,
"acc": 0.734375,
"acc_se": 0.03634097158237767,
"cut_share": 0.12109375,
"mean_tokens": 12056.03125
},
"math": {
"problems": 128,
"samples": 256,
"acc": 0.47265625,
"acc_se": 0.041245299835868436,
"cut_share": 0.44921875,
"mean_tokens": 22667.92578125
}
},
"run": {
"name": "dev/l15only",
"model": "qwen35-l15only",
"data": "dev256",
"n": 2,
"shards": 1,
"est": 11
},
"sampling": {
"temperature": 0.6,
"top_p": 0.95,
"top_k": 20,
"presence_penalty": 1.5,
"repetition_penalty": 1.0,
"max_tokens": 32768,
"thinking": true
},
"edit_note": {
"name": "l15only",
"spec": {
"about": "the released edit on layer 15 only",
"layers": {
"15": {
"alpha": 6
}
}
},
"built": {
"15": {
"alpha": 6.0,
"n_cols": 64,
"dir": "released",
"relative_change": 0.10635253041982651
}
},
"edited_sha256": {
"model.language_model.layers.15.mlp.down_proj.weight": "571c7435b3eedc782f4daa6b6e2cb5d86302624bb5893b087d53865585f7448a"
}
},
"tokens_total": 8889333,
"gpu_seconds": 1337,
"tok_per_s_per_gpu": 6886
}