Jonathan Tow
91 lines
2.8 KiB
JSON
91 lines
2.8 KiB
JSON
{
|
|
"results": {
|
|
"arc_challenge": {
|
|
"acc": 0.3848122866894198,
|
|
"acc_stderr": 0.014218371065251107,
|
|
"acc_norm": 0.4052901023890785,
|
|
"acc_norm_stderr": 0.014346869060229334
|
|
},
|
|
"arc_easy": {
|
|
"acc": 0.7319023569023569,
|
|
"acc_stderr": 0.009089526578213694,
|
|
"acc_norm": 0.6910774410774411,
|
|
"acc_norm_stderr": 0.009481048387761353
|
|
},
|
|
"boolq": {
|
|
"acc": 0.7030581039755351,
|
|
"acc_stderr": 0.007991418738281637
|
|
},
|
|
"hellaswag": {
|
|
"acc": 0.5553674566819359,
|
|
"acc_stderr": 0.004959094146471525,
|
|
"acc_norm": 0.7426807408882693,
|
|
"acc_norm_stderr": 0.004362633637374482
|
|
},
|
|
"lambada_openai": {
|
|
"ppl": 3.366151815984524,
|
|
"ppl_stderr": 0.06573496070440447,
|
|
"acc": 0.7418979235396856,
|
|
"acc_stderr": 0.006096490478492321
|
|
},
|
|
"openbookqa": {
|
|
"acc": 0.304,
|
|
"acc_stderr": 0.020591649571224925,
|
|
"acc_norm": 0.418,
|
|
"acc_norm_stderr": 0.022080014812228134
|
|
},
|
|
"piqa": {
|
|
"acc": 0.7845484221980413,
|
|
"acc_stderr": 0.009592463115658117,
|
|
"acc_norm": 0.8019586507072906,
|
|
"acc_norm_stderr": 0.009298209954776726
|
|
},
|
|
"sciq": {
|
|
"acc": 0.939,
|
|
"acc_stderr": 0.007572076091557423,
|
|
"acc_norm": 0.917,
|
|
"acc_norm_stderr": 0.00872852720607479
|
|
},
|
|
"siqa": {
|
|
"acc": 0.4242579324462641,
|
|
"acc_stderr": 0.011183502662341787,
|
|
"acc_norm": 0.4692937563971341,
|
|
"acc_norm_stderr": 0.011292714928103489
|
|
},
|
|
"truthfulqa_mc": {
|
|
"mc1": 0.2423500611995104,
|
|
"mc1_stderr": 0.01500067437357034,
|
|
"mc2": 0.36457253410937596,
|
|
"mc2_stderr": 0.0135567172376818
|
|
},
|
|
"winogrande": {
|
|
"acc": 0.6882399368587214,
|
|
"acc_stderr": 0.013018571197638537
|
|
}
|
|
},
|
|
"versions": {
|
|
"arc_challenge": 0,
|
|
"arc_easy": 0,
|
|
"boolq": 1,
|
|
"hellaswag": 0,
|
|
"lambada_openai": 0,
|
|
"openbookqa": 0,
|
|
"piqa": 0,
|
|
"sciq": 0,
|
|
"siqa": 0,
|
|
"truthfulqa_mc": 1,
|
|
"winogrande": 0
|
|
},
|
|
"config": {
|
|
"model": "gpt2",
|
|
"model_args": "use_fast=True,pretrained=stabilityai/stablelm-base-alpha-7b-v2,trust_remote_code=True,low_cpu_mem_usage=True,dtype=auto",
|
|
"num_fewshot": 0,
|
|
"batch_size": "2",
|
|
"batch_sizes": [],
|
|
"device": "cuda:2",
|
|
"no_cache": false,
|
|
"limit": null,
|
|
"bootstrap_iters": 100000,
|
|
"description_dict": {}
|
|
}
|
|
} |