QTIP Quantized Models
Collection
See https://github.com/Cornell-RelaxML/qtip
•
30 items
•
Updated
•
9
BF16:
arc_challenge
{'alias': 'arc_challenge', 'acc,none': 0.5332764505119454, 'acc_stderr,none': 0.014578995859605814, 'acc_norm,none': 0.5324232081911263, 'acc_norm_stderr,none': 0.014580637569995418}
arc_easy
{'alias': 'arc_easy', 'acc,none': 0.7558922558922558, 'acc_stderr,none': 0.008814322157999389, 'acc_norm,none': 0.6435185185185185, 'acc_norm_stderr,none': 0.009828046544504433}
hellaswag
{'alias': 'hellaswag', 'acc,none': 0.5852419836685919, 'acc_stderr,none': 0.004916733258140271, 'acc_norm,none': 0.6487751443935471, 'acc_norm_stderr,none': 0.00476377498183474}
piqa
{'alias': 'piqa', 'acc,none': 0.8003264417845484, 'acc_stderr,none': 0.009326942154519176, 'acc_norm,none': 0.7899891186071817, 'acc_norm_stderr,none': 0.00950335330581858}
This model:
{'alias': 'arc_challenge', 'acc,none': 0.5366894197952219, 'acc_stderr,none': 0.014572000527756989, 'acc_norm,none': 0.514505119453925, 'acc_norm_stderr,none': 0.014605241081370053}
arc_easy
{'alias': 'arc_easy', 'acc,none': 0.7462121212121212, 'acc_stderr,none': 0.008929657065808293, 'acc_norm,none': 0.6292087542087542, 'acc_norm_stderr,none': 0.009911292822056925}
hellaswag
{'alias': 'hellaswag', 'acc,none': 0.581557458673571, 'acc_stderr,none': 0.004922953651577666, 'acc_norm,none': 0.6630153355905198, 'acc_norm_stderr,none': 0.00471713572219413}
piqa
{'alias': 'piqa', 'acc,none': 0.8052230685527747, 'acc_stderr,none': 0.00924000669331772, 'acc_norm,none': 0.8030467899891186, 'acc_norm_stderr,none': 0.009278918898006383}
Base model
meta-llama/Llama-3.1-70B