thegovind's picture
Publish the weight-space programming labs
bb102c0 verified
Raw History Blame Contribute Delete
1.63 kB
{
"problems": 256,
"samples": 512,
"acc": 0.46875,
"acc_se": 0.026903627024228972,
"cut_share": 0.490234375,
"mean_tokens": 20728.71875,
"closed_share": 0.6171875,
"p_finish": 0.509765625,
"p_correct_given_finished": 0.9118773946360154,
"p_correct_given_cut": 0.00796812749003984,
"no_answer_finished_share": 0.0,
"loop_end_share_of_cut": 0.900398406374502,
"tail_dup8_cut_mean": 0.8424358565737052,
"median_tokens_correct": 8348.0,
"by_source": {
"gsm8k": {
"problems": 128,
"samples": 256,
"acc": 0.63671875,
"acc_se": 0.03538787977024572,
"cut_share": 0.296875,
"mean_tokens": 15575.28515625
},
"math": {
"problems": 128,
"samples": 256,
"acc": 0.30078125,
"acc_se": 0.034785197395756036,
"cut_share": 0.68359375,
"mean_tokens": 25882.15234375
}
},
"run": {
"name": "dev/l19only",
"model": "qwen35-l19only",
"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": "l19only",
"spec": {
"about": "the released edit on layer 19 only",
"layers": {
"19": {
"alpha": 6
}
}
},
"built": {
"19": {
"alpha": 6.0,
"n_cols": 64,
"dir": "released",
"relative_change": 0.0810554027557373
}
},
"edited_sha256": {
"model.language_model.layers.19.mlp.down_proj.weight": "b1e1949735dc682b9d7f18f4bc13eb07dce934aab85f302965d8dba403c788bc"
}
},
"tokens_total": 10613104,
"gpu_seconds": 1753,
"tok_per_s_per_gpu": 6219
}