Results

Quant Performance

Tasks Version Filter n-shot Metric Value Stderr
gsm8k 3 flexible-extract 5 exact_match 0.8135 ± 0.0107
strict-match 5 exact_match 0.8059 ± 0.0109
hellaswag 1 none 0 acc 0.6308 ± 0.0048
none 0 acc_norm 0.8256 ± 0.0038
humaneval 1 create_test 0 pass@1 0.4146 ± 0.0386
mbpp 1 none 3 pass_at_1 0.5860 ± 0.0220
piqa 1 none 0 acc 0.8194 ± 0.0090
none 0 acc_norm 0.8303 ± 0.0088
rte 1 none 0 acc 0.7870 ± 0.0246

Base Performance

Tasks Version Filter n-shot Metric Value Stderr
gsm8k 3 flexible-extract 5 exact_match 0.9242 ± 0.0073
strict-match 5 exact_match 0.9219 ± 0.0074
hellaswag 1 none 0 acc 0.6548 ± 0.0047
none 0 acc_norm 0.8496 ± 0.0036
humaneval 1 create_test 0 pass@1 0.4329 ± 0.0388
mbpp 1 none 3 pass_at_1 0.6800 ± 0.0209
piqa 1 none 0 acc 0.8270 ± 0.0088
none 0 acc_norm 0.8400 ± 0.0086
rte 1 none 0 acc 0.7220 ± 0.0270

Differences

Task Metric Quant Value Base Value Difference (Base - Quant)
gsm8k exact_match 0.8135 0.9242 +0.1107
gsm8k exact_match (strict) 0.8059 0.9219 +0.1160
hellaswag acc 0.6308 0.6548 +0.0240
hellaswag acc_norm 0.8256 0.8496 +0.0240
humaneval pass@1 0.4146 0.4329 +0.0183
mbpp pass_at_1 0.5860 0.6800 +0.0940
piqa acc 0.8194 0.8270 +0.0076
piqa acc_norm 0.8303 0.8400 +0.0097
rte acc 0.7870 0.7220 -0.0650
Downloads last month
16
Safetensors
Model size
8.08B params
Tensor type
I64
·
I32
·
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for cgg507/Valkyrie-v1-gptq-channel

Quantized
(29)
this model